Reactions to Apple Intelligence, which Apple unveiled at WWDC 2024, have ranged from curious to positive to underwhelmed, but whatever your views on the technology itself, a big talking point has been Apple’s emphasis on privacy, in contrast to some companies that have been offering generative AI products for some time.
Apple is putting privacy front and center with its AI offering and has been keen to talk about how Apple Intelligence – which will be integrated across iOS 18, iPadOS 18, and macOS Sequoia – would differ from its competitors by adopting a fresh approach to handling user information.
Craig Federighi, Apple’s Senior Vice President of Software Engineering, and the main presenter of the WWDC keynote, has been sharing more details about Apple Intelligence, and the company’s privacy-first approach.
Speaking to Fast Company, Federighi explained more about Apple’s overall AI ambitions, confirming that Apple is in agreement with other big tech companies that generative AI is the next big thing – as big a thing as the internet or microprocessors were when they first came about – and that we’re at the beginning of generative AI’s evolution.
(Image credit: Apple)
Apple's commitment to AI privacy
Federighi told Fast Company that Apple is aiming to “establish an entirely different bar” to other AI services and products when it comes to privacy. He reinforced the messaging in the WWDC keynote that the personal aspect of Apple Intelligence is foundational to it and that users’ information will be under their control. He also reiterated that Apple wouldn’t be able to access your information, even while its data centers are processing it.
The practical measures that Apple is taking to achieve this begin with its lineup of Apple M-series processors, which it claims will be able to run and process many AI tasks on-device, meaning your data won’t have to leave your system. For times when that local processing power is insufficient, the task at hand will be sent to dedicated custom-built Apple servers utilizing Private Cloud Compute (PCC), offering far more grunt for requests that need it – while being more secure than other cloud products in the same vein, Apple claims.
This will mean that your device will only send the minimum information required to process your requests, and Apple claims that its servers are designed in such a way that it’s impossible for them to store your data. This is apparently because after your request is processed and returned to your device, the information is ‘cryptographically destroyed’, and is never seen by anyone at Apple.
Apple has published a more in-depth security research blog post going into more detail about PCC, which, as noted at WWDC 2024, is a system available to independent security researchers, who can access Apple Intelligence servers in order to verify Apple’s privacy and security claims around PCC.
Apple wants AI to feel like a natural, almost unnoticeable part of its software, and the tech giant is clearly keen to win the trust of those who use its products and to differentiate its take on AI compared with that of rivals.
(Image credit: Apple)
More about ChatGPT and Apple Intelligence in China
Federighi also talks about Apple’s new partnership with OpenAI and the integration of ChatGPT into its operating systems. This is being done to give users access to industry-standard advanced models while reassuring users that ChatGPT isn’t what powers Apple Intelligence; the latter is exclusively driven by Apple’s own large language models (LLMs), which are totally distinct on Apple’s platforms, but you will be able to enlist ChatGPT for more complex requests.
ChatGPT is only ever invoked at the user’s request and with their permission, and before any requests are sent to ChatGPT you’ll have to confirm that you want to do this explicitly. Apple teamed up with OpenAI to give users this option because, according to Federighi, GPT-4o is “currently the best LLM out there for broad world knowledge.”
Apple is also considering expanding this concept to include other LLM makers in the future so that you might be able to choose from a variety of LLMs for your more demanding requests.
Federighi also talked about its plans for Apple Intelligence in China – the company’s second biggest market – and how the company is working to comply with regulations in the country to bring its most cutting-edge capabilities to all customers. This process is underway, but may take a while, as Federighi observed: “We don’t have timing to announce right now, but it’s certainly something we want to do.”
We’ll have to see how Apple Intelligence performs in practice, and if Apple’s privacy-first approach pays off. Apple has a strong track record when it comes to designing products and services that integrate so seamlessly that they become a part of our everyday lives, and it might very well be on track to continue building that reputation with Apple Intelligence.
Almost everyone in tech is investing heavily in artificial intelligence right now, and Google is among those most committed to an AI future. Project Astra, unveiled at Google I/O 2024, is a big part of that – and it could end up being one of Google's most important AI tools.
Astra is being billed as “a universal AI agent that is helpful in everyday life”. It's essentially something like a blending of Google Assistant and Google Gemini, with added features and supercharged capabilities for a natural, conversational experience.
Here, we're going to explain everything you need to know about Project Astra – how it works, what it can do, when you can get it, and how it might shape the future.
What is Project Astra?
In some ways, Project Astra isn't any different to the AI chatbots we've already got: you ask a question about what's in a picture, or about how to do something, or request some creative text to be generated, and Astra gets on with it.
What elevates this particular AI project is its multimodal functionality (the way text, images, video, and audio can all be combined), the speed that the bot works at, and how conversational it is. Google's aim, as we've already mentioned, is to create “a universal AI agent” that can do anything and understand everything.
Project Astra in action (Image credit: Google)
Think about the Hal 9000 bot in Kubrick's 2001: A Space Odyssey, or the Samantha assistant in the movie Her: talking to them is like talking to a human being, and there isn't much they can't do. (Both those AIs eventually got too big for their creators to control, but let's ignore that for the time being.)
Project Astra has been built to understand context and to take actions, to be able to work in real time, and to remember conversations from the past. From the demos we've seen so far, it works on phones and on smart glasses, and is powered by the Google Gemini AI models – so it may eventually be part of the Gemini app, rather than something that's separate and standalone.
When is Project Astra coming out?
Project Astra is in its early stages: this isn't something that's going to be available to the masses for a few months at least. That said, Google says that “some of these agent capabilities will come to Google products like the Gemini app later this year”, so it looks as though elements of Astra will appear gradually in Google's apps as we go through 2024.
When we were given some hands-on time with Project Astra at I/O 2024, these sessions were limited to four minutes each – so that gives you some idea of how far away this is from being something that anyone, anywhere can make use of. What's more, the Astra kit didn't look particularly portable, and the Google reps were careful to refer to it as a prototype.
Taking all that together, we get the impression that some of the Project Astra tricks we've seen demoed might appear in the Google Gemini app sooner rather than later. At the same time, the full Astra experience – perhaps involving some dedicated hardware – is probably not going to be rolling out until 2025 at the earliest.
Now that Google has shared what Project Astra is and what it's capable of, it's likely that we're going to hear a whole lot more about it in the months ahead. Bear in mind that ChatGPT and Dall-E developer OpenAI is busy pushing out major upgrades of its own, and Google isn't going to want to be left behind.
What can I do with Project Astra?
One of Google's demos shows Astra running on a phone, using its camera input and talking naturally to a user: it's asked to flag up something in view that can play sounds, and correctly identifies a speaker. When an arrow is drawn on screen, Astra then recognizes and talks about the speaker component highlighted by the arrow.
In another demo, we see Astra correctly identifying world landmarks from drawings in a sketchbook. It's also able to remember the order of objects in a list, identify a neighborhood from an image, understand the purpose of sections of code that are shown to it, and solve math problems that are written out.
There's a lot of emphasis on recognizing objects, drawings, text, and more through a camera system – while at the same time understanding human speech and generating appropriate responses. This is the multimodal part of Project Astra in action, which makes it a step up from what we already have – with improvements in caching, recording, and processing key to the real time responsiveness.
In our hands-on time with Project Astra, we were able to get it to tell a story based on objects that we showed to the camera – and adapt the story as we went on. Further down the line, it's not difficult to imagine Astra applying these smarts as you explore a city on vacation, or solve a physics problem on a whiteboard, or provide detailed information about what's being shown in a sports game.
Which devices will include Project Astra?
In the demonstrations of Project Astra that Google has shown off so far, the AI is running on an unidentified smartphone and an unidentified pair of smart glasses – suggesting that we might not have heard the last of Google Glass yet.
Google has also hinted that Project Astra is going to be coming to devices with other form factors. We've already mentioned the Her movie, and it's well within the realms of possibility that we might eventually see the Astra bot built into wireless earbuds (assuming they have a strong enough Wi-Fi connection).
Expect to see Project Astra turn up on Pixel phones, eventually (Image credit: Future / Philip Berne)
In the hands-on area that was set up at Google I/O 2024, Astra was powered through a large camera, and could only work with a specific set of objects as props. Clearly, any device that runs Astra's impressive features is going to need a lot of on-board processing power, or a very quick connection to the cloud, in order to keep up the real-time conversation that's core to the AI.
As time goes on and technology improves, though, these limitations should slowly begin to be overcome. The next time we hear something major about Project Astra could be around the time of the launch of the Google Pixel 9 in the last few months of 2024; Google will no doubt want to make this the most AI-capable smartphone yet.
Apple announced the M4 chip, a powerful new upgrade that will arrive in next-generation iPad (and, further down the line, the best Macbooks and Macs). You can check out our beat-by-beat coverage of the Apple event, but one element of the presentation has left some users confused: what exactly does TOPS mean?
TOPS is an acronym for 'trillion operations per second', and is essentially a hardware-specific measure of AI capabilities. More TOPS means faster on-chip AI performance, in this case the Neural Engine found on the Apple M4 chip.
The M4 chip is capable of 38 TOPS – that's 38,000,000,000,000 operations per second. If that sounds like a staggeringly massive number, well, it is! Modern neural processing units (NPUs) like Apple's Neural Engine are advancing at an incredibly rapid rate; for example, Apple's own A16 Bionic chip, which debuted in the iPhone 14 Pro less than two years ago, offered 17 TOPS.
Apple's new chip isn't even the most powerful AI chip about to hit the market – Qualcomm's upcoming Snapdragon X Elite purportedly offers 45 TOPS, and is expected to land in Windows laptops later this year.
How is TOPS calculated?
The processes by which we measure AI performance are still in relative infancy, but TOPS provides a useful and user-accessible metric for discerning how 'good' at handling AI tools a given processor is.
I'm about to get technical, so if you don't care about the mathematics, feel free to skip ahead to the next section! The current industry standard for calculating TOPS is TOPS = 2 × MAC unit count × Frequency / 1 trillion. 'MAC' stands for multiply-accumulate; a MAC operation is basically a pair of calculations (a multiplication and an addition) that are run by each MAC unit on the processor once every clock cycle, powering the formulas that make AI models function. Every NPU has a set number of MAC units determined by the NPU's microarchitecture.
'Frequency' here is defined by the clock speed of the processor in question – specifically, how many cycles it can process per second. It's a common metric also used in CPUs, GPUs, and other components, essentially denoting how 'fast' the component is.
So, to calculate how many operations per second an NPU can handle, we simply multiply the MAC unit count by 2 for our number of operations, then multiply that by the frequency. This gives us an 'OPS' figure, which we then divide by a trillion to make it a bit more palatable (and kinder on your zero key when typing it out).
Simply put, more TOPS means better, faster AI performance.
Adobe’s Firefly generative AI tool can be hardware-accelerated by your device’s NPU. (Image credit: Adobe)
Why is TOPS important?
TOPS is, in the simplest possible terms, our current best way to judge the performance of a device for running local AI workloads. This applies both to the industry and the wider public; it's a straightforward number that lets professionals and consumers immediately compare the baseline AI performance of different devices.
TOPS is only applicable for on-device AI, meaning that cloud-based AI tools (like the internet's favorite AI bot, ChatGPT) don't typically benefit from better TOPS. However, local AI is becoming more and more prevalent, with popular professional software like the Adobe Creative Cloud suite starting to implement more AI-powered features that depend on the capabilities of your device.
It should be noted that TOPS is by no means a perfect metric. At the end of the day, it's a theoretical figure derived from hardware statistics and can differ greatly from real-world performance. Factors such as power availability, thermal systems, and overclocking can impact the actual speed at which an NPU can run AI workloads.
To that end, though, we're now starting to see AI benchmarks crop up, such as Procyon AI from UL Benchmarks (makers of the popular 3DMark and PCMark benchmarking programs). These can provide a much more realistic idea of how well a You can expect to see TechRadar running AI performance tests as part of our review benchmarking in the near future!
ChatGPT maker OpenAI has now unveiled Sora, its artificial intelligence engine for converting text prompts into video. Think Dall-E (also developed by OpenAI), but for movies rather than static images.
It's still very early days for Sora, but the AI model is already generating a lot of buzz on social media, with multiple clips doing the rounds – clips that look as if they've been put together by a team of actors and filmmakers.
Here we'll explain everything you need to know about OpenAI Sora: what it's capable of, how it works, and when you might be able to use it yourself. The era of AI text-prompt filmmaking has now arrived.
OpenAI Sora release date and price
In February 2024, OpenAI Sora was made available to “red teamers” – that's people whose job it is to test the security and stability of a product. OpenAI has also now invited a select number of visual artists, designers, and movie makers to test out the video generation capabilities and provide feedback.
“We're sharing our research progress early to start working with and getting feedback from people outside of OpenAI and to give the public a sense of what AI capabilities are on the horizon,” says OpenAI.
In other words, the rest of us can't use it yet. For the time being there's no indication as to when Sora might become available to the wider public, or how much we'll have to pay to access it.
(Image credit: OpenAI)
We can make some rough guesses about timescale based on what happened with ChatGPT. Before that AI chatbot was released to the public in November 2022, it was preceded by a predecessor called InstructGPT earlier that year. Also, OpenAI's DevDay typically takes place annually in November.
It's certainly possible, then, that Sora could follow a similar pattern and launch to the public at a similar time in 2024. But this is currently just speculation and we'll update this page as soon as we get any clearer indication about a Sora release date.
As for price, we similarly don't have any hints of how much Sora might cost. As a guide, ChatGPT Plus – which offers access to the newest Large Language Models (LLMs) and Dall-E – currently costs $ 20 (about £16 / AU$ 30) per month.
But Sora also demands significantly more compute power than, for example, generating a single image with Dall-E, and the process also takes longer. So it still isn't clear exactly how well Sora, which is effectively a research paper, might convert into an affordable consumer product.
What is OpenAI Sora?
You may well be familiar with generative AI models – such as Google Gemini for text and Dall-E for images – which can produce new content based on vast amounts of training data. If you ask ChatGPT to write you a poem, for example, what you get back will be based on lots and lots of poems that the AI has already absorbed and analyzed.
OpenAI Sora is a similar idea, but for video clips. You give it a text prompt, like “woman walking down a city street at night” or “car driving through a forest” and you get back a video. As with AI image models, you can get very specific when it comes to saying what should be included in the clip and the style of the footage you want to see.
To get a better idea of how this works, check out some of the example videos posted by OpenAI CEO Sam Altman – not long after Sora was unveiled to the world, Altman responded to prompts put forward on social media, returning videos based on text like “a wizard wearing a pointed hat and a blue robe with white stars casting a spell that shoots lightning from his hand and holding an old tome in his other hand”.
How does OpenAI Sora work?
On a simplified level, the technology behind Sora is the same technology that lets you search for pictures of a dog or a cat on the web. Show an AI enough photos of a dog or cat, and it'll be able to spot the same patterns in new images; in the same way, if you train an AI on a million videos of a sunset or a waterfall, it'll be able to generate its own.
Of course there's a lot of complexity underneath that, and OpenAI has provided a deep dive into how its AI model works. It's trained on “internet-scale data” to know what realistic videos look like, first analyzing the clips to know what it's looking at, then learning how to produce its own versions when asked.
So, ask Sora to produce a clip of a fish tank, and it'll come back with an approximation based on all the fish tank videos it's seen. It makes use of what are known as visual patches, smaller building blocks that help the AI to understand what should go where and how different elements of a video should interact and progress, frame by frame.
Sora starts messier, then gets tidier (Image credit: OpenAI)
Sora is based on a diffusion model, where the AI starts with a 'noisy' response and then works towards a 'clean' output through a series of feedback loops and prediction calculations. You can see this in the frames above, where a video of a dog playing in the show turns from nonsensical blobs into something that actually looks realistic.
And like other generative AI models, Sora uses transformer technology (the last T in ChatGPT stands for Transformer). Transformers use a variety of sophisticated data analysis techniques to process heaps of data – they can understand the most important and least important parts of what's being analyzed, and figure out the surrounding context and relationships between these data chunks.
What we don't fully know is where OpenAI found its training data from – it hasn't said which video libraries have been used to power Sora, though we do know it has partnerships with content databases such as Shutterstock. In some cases, you can see the similarities between the training data and the output Sora is producing.
What can you do with OpenAI Sora?
At the moment, Sora is capable of producing HD videos of up to a minute, without any sound attached, from text prompts. If you want to see some examples of what's possible, we've put together a list of 11 mind-blowing Sora shorts for you to take a look at – including fluffy Pixar-style animated characters and astronauts with knitted helmets.
“Sora can generate videos up to a minute long while maintaining visual quality and adherence to the user’s prompt,” says OpenAI, but that's not all. It can also generate videos from still images, fill in missing frames in existing videos, and seamlessly stitch multiple videos together. It can create static images too, or produce endless loops from clips provided to it.
It can even produce simulations of video games such as Minecraft, again based on vast amounts of training data that teach it what a game like Minecraft should look like. We've already seen a demo where Sora is able to control a player in a Minecraft-style environment, while also accurately rendering the surrounding details.
OpenAI does acknowledge some of the limitations of Sora at the moment. The physics don't always make sense, with people disappearing or transforming or blending into other objects. Sora isn't mapping out a scene with individual actors and props, it's making an incredible number of calculations about where pixels should go from frame to frame.
In Sora videos people might move in ways that defy the laws of physics, or details – such as a bite being taken out of a cookie – might not be remembered from one frame to the next. OpenAI is aware of these issues and is working to fix them, and you can check out some of the examples on the OpenAI Sora website to see what we mean.
Despite those bugs, further down the line OpenAI is hoping that Sora could evolve to become a realistic simulator of physical and digital worlds. In the years to come, the Sora tech could be used to generate imaginary virtual worlds for us to explore, or enable us to fully explore real places that are replicated in AI.
How can you use OpenAI Sora?
At the moment, you can't get into Sora without an invite: it seems as though OpenAI is picking out individual creators and testers to help get its video-generated AI model ready for a full public release. How long this preview period is going to last, whether it's months or years, remains to be seen – but OpenAI has previously shown a willingness to move as fast as possible when it comes to its AI projects.
Based on the existing technologies that OpenAI has made public – Dall-E and ChatGPT – it seems likely that Sora will initially be available as a web app. Since its launch ChatGPT has got smarter and added new features, including custom bots, and it's likely that Sora will follow the same path when it launches in full.
Before that happens, OpenAI says it wants to put some safety guardrails in place: you're not going to be able to generate videos showing extreme violence, sexual content, hateful imagery, or celebrity likenesses. There are also plans to combat misinformation by including metadata in Sora videos that indicates they were generated by AI.
Google has been busy revamping its AI offerings, renaming Bard to Gemini, pushing out a dedicated Android app, and lots more besides. There's also now a paid tier for Google's generative AI engine for the first time, which means another digital subscription for you to weigh up.
You can read our Google Gemini explained explainer for a broad overview of Google's AI tools. But here we'll be breaking down the Google Gemini Advanced features that come as part of the new Google One AI Premium tier.
We'll be exploring how much this new cloud tier costs, plus all the AI features and benefits it brings, so you can decide whether or not you'd like to sign up. It's been added as one of the Google One plans, so you get some digital storage in the cloud included, too. Here's how Google One AI Premium is shaping up so far…
Google One AI Premium: price and availability
The Google One AI Premium plan is available to buy now and will cost you $ 19.99 / £18.99 / AU$ 32.99 a month. Unlike some other Google One plans, you can't pay annually to get a discount on the overall price, but you can cancel whenever you like.
At the time of writing, Google is offering free two-month trials of Google One AI Premium, so you won't have to pay anything for the first two months. You can sign up and compare plans on the Google One site.
Google One AI Premium: features and benefits
First of all, you get 2TB of storage to use across your Google services: Gmail, Google Drive, and Google Photos. If you've been hitting the limits of the free storage plan – a measly 15GB – then that's another reason to upgrade.
You'll notice a variety of other Google One plans are available, offering storage from 2TB to 30TB, but it's only the Google One AI Premium plan that comes with all of the Gemini Advanced features.
Besides the actual storage space, all Google One plans include priority support, 10% back in the Google Store, extra Google Photos editing features (including Magic Eraser), a dark web monitoring service that'll look for any leaks of your personal information, and use of the Google One VPN.
Google Gemini Advanced on the web (Image credit: Google)
It's the AI features that you're here for though, and the key part of Google One AI Premium is that you get access to Gemini Advanced: that means the “most capable” version of Google's Gemini model, known as Ultra 1.0. You can think of it a bit like paying for ChatGPT Plus compared to sticking on the free ChatGPT plan.
Google describes Gemini Ultra 1.0 as offering “state-of-the-art performance” that's capable of handling “highly complex tasks” – tasks that can involve text, images, and code. Longer conversations are possible with Gemini Advanced, and it understands context better too. If you want the most powerful AI that Google has to offer, this is it.
A premium subscription will supercharge the Gemini app (Image credit: Google)
“The largest model Ultra 1.0 is the first to outperform human experts on MMLU (massive multitask language understanding), which uses a combination of 57 subjects — including math, physics, history, law, medicine and ethics — to test knowledge and problem-solving abilities,” writes Google CEO Sundar Pichai.
The dedicated Google Gemini app for Android, and the Gemini features built into the Google app for iOS, are available to everyone, whether they pay for a subscription or not – and it's the same with the web interface. However, if you're on the premium plan, you'll get the superior Ultra 1.0 model in all these places.
By the way, a standard 2TB Google One plan – with everything from the photo editing tricks to the VPN, but without the AI – will cost you $ 9.99 / £7.99 / AU$ 19.99 a month, so you're effectively paying $ 10 / £11 / AU$ 13 for Gemini Advanced.
An example of Google Gemini in Gmail (Image credit: Google)
Gemini integration with Google's productivity apps – including Gmail, Google Docs, Google Meet, and Google Slides – is going to be “available soon”, Google says, and when it does become available, you'll get it as part of a Google One AI Premium plan. It'll give you help in composing your emails, designing your slideshows, and so on.
This is a rebranding of the Duet AI features that Google has previously rolled out for users of its apps, and it's now known as Gemini for Workspace. Whether you're an individual or a business user though, you'll be able to get these integrated AI tools if you sign up for the Google One AI Premium plan.
So there you have it: beyond the standard 2TB Google One plan, the main takeaway is that you get access to the latest and greatest Gemini AI features from Google, and the company is promising that there will be plenty more on the way in the future, too.
Google One AI Premium early verdict
On one hand, Google's free two-month trial of the One AI Premium Plan (which contains Gemini Advanced) feels like a no-brainer for those who want to tinker with some of the most powerful AI tools available right now. As long as you're fairly disciplined about canceling unwanted free trials, of course.
But it's also still very early days for Gemini Advanced. We haven't yet been able to put it through its paces or compare it to the likes of ChatGPT Plus. Its integration with Google's productivity apps is also only “available soon”, so it's not yet clear when that will happen.
(Image credit: Google)
If you want to deep dive into the performance of Google's latest AI models – including Gemini Advanced – you can read the company's Gemini benchmarking report. Some lucky testers like AI professor Ethan Mollick have also been tinkering with Gemini Advanced for some time after getting advanced access.
The early impressions seem to be that Gemini Advanced is shaping up to be a GPT-4 class AI contender that's capable of competing with ChatGPT Plus for demanding tasks like coding and advanced problem-solving. It also promises to integrate nicely with Google's apps. How well it does that in reality is something we'll have to wait a little while to find out, but that free trial is there for early adopters who want to dive straight in.
Google has been a sleeping AI giant, but this week it finally woke up. Google Gemini is here and it's the tech giant's most powerful range of AI tools so far. But Gemini is also, in true Google style, really confusing, so we're here to quickly break it all down for you.
Gemini is the new umbrella name for all of Google's AI tools, from chatbots to voice assistants and full-blown coding assistants. It replaces both Google Bard – the previous name for Google's AI chatbot – and Duet AI, the name for Google's Workspace-oriented rival to CoPilot Pro and ChatGPT Plus.
But this is also way more than just a rebrand. As part of the launch, Google has released a new free Google Gemini app for Android (in the US, for now. For the first time, Google is also releasing its most powerful large language model (LLM) so far called Gemini Ultra 1.0. You can play with that now as well, if you sign up for its new Google One AI Premium subscription (more on that below).
This is all pretty head-spinning stuff, and we haven't even scratched the surface of what you can actually do with these AI tools yet. So for a quick fast-charge to get you up to speed on everything Google Gemini, plug into our easily-digestible explainer below…
1. Gemini replaces Google Bard and Duet AI
In some ways, Google Gemini makes things simpler. It's the new umbrella name for all of Google's AI tools, whether you're on a smartphone or desktop, or using the free or paid versions.
Gemini replaces Google Bard (the previous name for Google's “experimental” AI chatbot) and Duet AI, the collection of work-oriented tools for Google Workspace. Looking for a free AI helper to make you images or redraft emails? You can now go to Google Gemini and start using it with a standard Google account.
But if you want the more powerful Gemini Advanced AI tools – and access to Google's newest Gemini Ultra LLM – you'll need to pay a monthly subscription. That comes as part of a Google One AI Premium Plan, which you can read more about below.
To sum up, there are three main ways to access Google Gemini:
As we mentioned above, Google has launched a new free Gemini app for Android. This is rolling out in the US now and Google says it'll be “fully available in the coming weeks”, with more locations to “coming soon”. Google is known for having a broad definition of “soon”, so the UK and EU may need to be patient.
There's going to be a similar rollout for iOS and iPhones, but with a different approach. Rather than a separate standalone app, Gemini will be available in the Google app.
The Android app is a big deal in particular because it'll let you set Gemini as your default voice assistant, replacing the existing Google Assistant. You can set this during the app's setup process, where you can tap “I agree” for Gemini to “handle tasks on your phone”.
Do this and it'll mean that whenever you summon a voice assistant on your Android phone – either by long-pressing your home button or saying “Hey Google” – you'll speak to Gemini rather than Google Assistant. That said, there is evidence that you may not want to do that just yet…
3. You may want to stick with Google Assistant (for now)
(Image credit: Google)
The Google Gemini app has only been out for a matter of days – and there are early signs of teething issues and limitations when it comes to using Gemini as your voice assistant.
The Play Store is filling up with complaints stating that Gemini asks you to tap 'submit' even when using voice commands and that it lacks functionality compared to Assistant, including being unable to handle hands-free reminders, home device control and more. We've also found some bugs during our early tests with the app.
Fortunately, you can switch back to the old Google Assistant. To do that, just go the Gemini app, tap your Profile in the top-right corner, then go to Settings > Digital assistants from Google. In here you'll be able to choose between Gemini and Google Assistant.
Sissie Hsiao (Google's VP and General Manager of Gemini experiences) claims that Gemini is “an important first step in building a true AI assistant – one that is conversational, multimodal and helpful”. But right now, it seems that “first step” is doing a lot of heavy lifting.
4. Gemini is a new way to quiz Google's other apps
(Image credit: Google)
Like the now-retired Bard, Gemini is designed to be a kind of creative co-pilot if you need help with “writing, brainstorming, learning, and more”, as Google describes it. So like before, you can ask it to tell you a joke, rewrite an email, help with research and more.
As always, the usual caveats remain. Google is still quite clear that “Gemini will make mistakes” and that, even though it's improving by the day, Gemini “can provide inaccurate information, or it can even make offensive statements”.
This means its other use case is potentially more interesting. Gemini is also a new way to interact with Google's other services like YouTube, Google Maps and Gmail. Ask it to “suggest some popular tourist sites in Seattle” and it'll show them in Google Maps.
Another example is asking it to “find videos of how to quickly get grape juice out of a wool rug”. This means Gemini is effectively a more conversational way to interact with the likes of YouTube and Google Drive. It can also now generate images, which was a skill Bard learnt last week before it was renamed.
5. The free version of Gemini has limitations
(Image credit: Future)
The free version of Gemini (which you access in the Google Gemini app on Android, in the Google app on iOS, or on the Gemini website) has quite a few limitations compared to the subscription-only Gemini Advanced.
This is partly because it's based on a simpler large language model (LLM) called Gemini Pro, rather than Google's new Gemini Ultra 1.0. Broadly speaking, the free version is less creative, less accurate, unable to handle multi-step questions, can't really code and has more limited data-handling powers.
This means the free version is best for basic things like answering simple questions, summarizing emails, making images, and (as we discussed above) quizzing Google's other services using natural language.
Looking for an AI assistant that can help with advanced coding, complex creative projects, and also work directly within Gmail and Google Docs? Google Gemini Advanced could be more up your street, particularly if you already subscribe to Google One…
6. Gemini Advanced is tempting for Google One users
The subscription-only Gemini Advanced costs $ 19.99 / £18.99 / AU$ 32.99 per month, although you can currently get a two-month free trial. Confusingly, you get Advanced by paying for a new Google One AI Premium Plan, which includes 2TB of cloud storage.
This means Gemini Advanced is particularly tempting if you already pay for a Google One cloud storage plan (or are looking to sign up for it anyway). With a 2TB Google One plan already costing $ 9.99 / £7.99 / AU$ 12.49 per month, that means the AI features are effectively setting you back an extra $ 10 / £11 / AU$ 20 a month.
There's even better news for those who already have a Google One subscription with 5TB of storage or more. Google says you can “enjoy AI Premium features until July 21, 2024, at no extra charge”.
This means that Google, in a similar style to Amazon Prime, is combining its subscriptions offerings (cloud storage and its most powerful AI assistant) in order to make them both more appealing (and, most likely, more sticky too).
7. The Gemini app could take a little while to reach the UK and EU
(Image credit: Future)
While Google has stated that the Gemini Android app is “coming soon” to “more countries and languages”, it hasn't given any timescale for when that'll happen – and a possible reason for the delay is that it's waiting for the EU AI Act to become clearer.
Sissie Hsiao (Google's VP and General Manager of Gemini experiences) told the MIT Technology Review “we’re working with local regulators to make sure that we’re abiding by local regime requirements before we can expand.”
While that sounds a bit ominous, Hsiao added that “rest assured, we are absolutely working on it and I hope we’ll be able to announce expansion very, very soon.” So if you're in the UK or EU, you'll need to settle for tinkering with the website version for now.
Given the early reviews of the Google Gemini Android app, and its inconsistencies as a Google Assistant replacement, that might well be for the best anyway.
We were first introduced to the Rabbit R1 in January 2024, at CES 2024, but what exactly is it? The charming sidekick (designed by Teenage Engineering) is promising to take pocket gadgets to the next level – offering something like a smartphone, but with an intuitive, unified, AI-driven interface that means you (theoretically, at least) need to interact with individual apps and websites.
If you're curious about the Rabbit R1 and the ways in which it might change the course of personal computing – or at least show us how next-gen smartphone voice assistants might work – we've gathered together everything you need to know about it here. From what it costs and how it works, to the AI engine driving the R1 experience, all the details of this potentially revolutionary device are here.
The first batches of the Rabbit R1 are due to start shipping to users later in 2024, although it seems availability is going to be rather limited to begin with – so you might have to wait a while to get your very own Rabbit R1.
Rabbit R1: one-minute overview
The r1 (Image credit: Rabbit)
The Rabbit R1 is a lot like a phone in terms of its looks, and in some of its features: it has a camera and a SIM card slot, and it supports Wi-Fi and Bluetooth. What's different, and what makes the Rabbit R1 special, is the interface: instead of a grid of apps, you get an AI assistant that talks to your favorite apps and does everything for you.
For example, you could get the R1 to research a holiday destination and book flights to it, or queue up a playlist of your favorite music, or book you a cab. In theory, you can do anything you can already do on your phone, just by asking. That said, there remain a lot of questions over exactly how it works and protects your privacy in the way it describes.
We've seen next-gen personal assistants depicted in movies like Her, and the R1 is trying to make that a reality – leveraging the latest AI capabilities to replace the traditional smartphone interface with something a lot more intuitive and slick.
Introducing r1. Watch the keynote. Order now: https://t.co/R3sOtVWoJ5 #CES2024 pic.twitter.com/niUmjFvKvEJanuary 9, 2024
See more
Another way to think about the Rabbit R1 is as an evolution of the Amazon Echo, Google Nest, and Apple HomePod smart speakers. The voice-controlled digital assistants on these devices can do some rudimentary tasks – like check the weather or play music – but the R1 wants to go way beyond what they're capable of.
Rabbit says the R1 is “the future of human-machine interfaces”, and you can check out its pitch for the device in its very Apple-flavored CES 2024 keynote below.
Rabbit r1: release date and price
The first batch of 10,000 units of the Rabbit R1 were made available to preorder at the same time as the device was announced at CES, on January 9, 2024. Those units quickly sold out, as did a second batch of 10,000 units made available shortly after.
Rabbit says that the people who got their preorders in should start having their devices shipped to them in March and April 2024. At the time of writing, there's no indication yet when another batch of units will be made available to preorder, or when we might see the r1 go on sale more widely.
What we do know is that the price of the Rabbit R1 starts at $ 199, which works out at around £155 / AU$ 300. To begin with, the Rabbit R1 is available to order in the US, Canada, the UK, Denmark, France, Germany, Ireland, Italy, Netherlands, Spain, Sweden, South Korea, and Japan, from the Rabbit website.
What's more, unlike rival AI devices such as the Humane AI Pin, there's no ongoing subscription fee that you have to pay out.
Rabbit r1: hardware
The r1 comes in a distinctive color (Image credit: Rabbit)
The Rabbit r1 is square, and bright orange, and comes with a 2.88-inch color touchscreen on the front. It's quite a compact device, almost small enough to fit in the palm of a hand, and it weighs in at 115 grams (about 4 oz). There's only one design, for now – you can't pick this up in multiple colors.
We know there's a far-field mic embedded in the R1, as well as built-in speakers. There's an integrated 360-degree camera here too, which is apparently called the Rabbit Eye. You can interact with elements by touching the screen, and there's an analog scroll wheel at the side of the device as well, if you need it.
The r1 camera (Image credit: Rabbit)
On the right of the Rabbit R1 is a push-to-talk button, which you make use of whenever you want to talk to the AI assistant on the device. There's no need for any “hey Google” or “hey Siri” wake command, and it also means the assistant doesn't have to be constantly listening out for your voice. Double-tapping the button activates the on-board camera.
Under the hood we've got a 2.3GHz MediaTek Helio processor, and Rabbit says the device offers “all day” battery life. That battery can be charged with a USB-C charge cable and power adapter, but it's worth bearing in mind that these aren't included in the box, so you'll have to use ones you've already got.
Rabbit R1: software
With its bright orange casing, the Rabbit r1 looks kind of cute, but it's the software that really makes it stand out. If you've used something like ChatGPT or Google Bard already, then this is something similar: Rabbit OS is fronted by an AI chatbot, capable of both answering questions and performing tasks.
In the CES keynote demo, Rabbit founder and CEO Jesse Lyu showed the R1 answering philosophical questions, checking stock prices, looking up information about movies, playing music on Spotify, booking an Uber, ordering a pizza, and planning a vacation (complete with flights and hotel reservations).
The r1 runs Rabbit OS (Image credit: Rabbit)
To get some of this working, you need to connect the Rabbit OS with your various apps and services, which can be done through a web portal. From the demo we've seen, it looks as though Spotify, Apple Music, YouTube Music, Expedia, Uber, eBay, and Amazon will be some of the services you can connect to.
Rabbit is keen to emphasize that it doesn't store any of your login details or track you in any way – it simply connects you to the apps you need – though the specifics of how it does this via the cloud are still unclear.
Rabbit's privacy page gives us a few more details, stating that “when you interact with rabbit OS, you will be assigned a dedicated and isolated environment on our cloud for your own LAM [large action model]. When our rabbits perform tasks for you, they will use your own accounts that you have securely granted us control over through our rabbit hole web portal”.
It also adds that “we do not store your passwords for these services. Rabbits will ask for permission and clarification during the execution of any tasks, especially those involving sensitive actions such as payments.” Exactly how Rabbit provides each user with a “dedicated and isolated environment” in its cloud isn't yet clear, but we should find out more once it lands with its first early adopters.
We've also been told the R1 can handle communication, real-time translation, and analyze images taken with the camera – show the R1 what's in your fridge, for example, and it could up with a dish you can cook.
The Rabbit R1 promises speedy responses too, quicker than you'd get with other generative AI bots. You can converse with the R1 as you would with Siri or Google Assistant, or you can bring up an on-screen keyboard by shaking the device. It calls its on-board AI a Large Action Model or LAM, similar to a Large Language Model or LLM (familiar from bots like ChatGPT), but with a lot more agency.
The r1 wants to take over multiple phone tasks (Image credit: Rabbit)
On top of all this, Rabbit says you can teach the R1 new skills. So, if you showed it how to go online and order your groceries for you, the next time it would be able to do that all by itself. In the CES demo, we saw the Rabbit R1 learning how to create AI images through Midjourney, and then replicating the process on its own.
Interestingly, Rabbit says it doesn't want the R1 to replace your phone – it wants to work alongside it. The R1 can't, for example, browse YouTube, check social media or let you organize your email (at least not yet), so it would seem that the humble smartphone will be sticking around for a while yet.
While some of the specifics about how the Rabbit R1 works and interacts with your favorite apps and services remain unclear, it's undoubtedly one of the most exciting pieces of AI hardware so far – as shown by the rapid sell-out of its early stock. We'll bring you more first impressions as soon as we've got our hands on one of 2024's early tech stars.
We’ve just got the first big update for macOS Sonoma (Apple’s latest operating system for Macs and MacBooks, which was released in September).
The Sonoma 14.1. update is available for all Mac users running macOS Sonoma, and can be downloaded and installed through the Software Update section found in System Settings.
If you’re not running macOS Sonoma, you’re not being left out, as Apple also released updates for older devices and operating systems, macOS Ventura 13.6.1 and macOS Monterey 12.7.1, which include many of the security fixes that macOS Sonoma 14.1 has.
The macOS Sonoma 14.1 update brings some new features to a range of apps, including a new warranty section which details your AppleCare+ plan (if you have one) and the status of your coverage (including for connected devices like AirPods and Beats headphones), along with new sections in the Apple Music app allowing you to add your favorite songs, albums, and playlists.
This isn’t a massive update, and seems almost like routine maintenance with some new additions, so there’s still plenty of room for improvement for macOS Sonoma, which is a decent operating system – but still not perfect. Some users are reporting buggy performance while using macOS Sonoma, although not all performance issues are Apple’s fault. That said, it seems like this update at least shows that Apple is aware of user feedback, and is working to improve the OS.
(Image credit: Apple)
What's coming next down the Apple pipeline
Hopefully, we won’t have long to wait for more improvements, as AppleInsider reports that macOS Sonoma 14.2’s developer beta has already been released to testers. If you would like to try this even newer version of macOS Sonoma, you’ll be able to grab it once the public beta version is released via the Apple Beta Software Program. This is only recommended for those willing to experiment with their devices, so we don’t recommend installing the beta on devices used for critical activities.
We recently learned that Apple has been tripling down on its AI efforts, and I think users are eager to see what this means for the company’s devices, such as the best MacBooks and Macs. Considering that Apple has been thought of as behind the curve in the recent round of the AI game, with competitors like Microsoft partnering with OpenAI and Amazon partnering with Anthropic (a rival of OpenAI working on innovative generative AI like its own AI chatbot, Claude), many people feel Apple needs to start showing off its AI products soon – maybe even in a future update for macOS Sonoma.
The Apple VR headset is getting close to its rumored arrival at WWDC 2023 on June 5 – and the mixed-reality wearable is expected be launched alongside an exciting new operating system, likely called xrOS.
What is xrOS? We may now be approaching iOS 17, iPadOS 16 and macOS 13 Ventura on Apple's other tech, but the Apple VR headset – rumored to be called the Apple Reality One – is expected to debut the first version of a new operating system that'll likely get regular updates just like its equivalents on iPhone, iPad and Mac.
The latest leaks suggest that Apple has settled on the xrOS name for its AR/VR headset, but a lot of questions remain. For example, what new things might xrOS allow developers (and us) to do in mixed reality compared to the likes of iOS? And will xrOS run ports of existing Apple apps like Freeform?
Here's everything we know so far about xrOS and the kinds of things it could allow Apple's mixed-reality headset to do in both augmented and virtual reality.
xrOS release date
It looks likely that Apple will launch its new xrOS operating system, alongside its new AR/VR headset, at WWDC 2023 on June 5. If you're looking to tune in, the event's keynote is scheduled to kick off at 10am PT / 1pm ET / 6pm BST (or 3am ACT on June 6).
This doesn't necessarily mean that a final version of xrOS will be released on that day. A likely scenario is that Apple will launch an xrOS developer kit to allow software makers to develop apps and experiences for the new headset.
Yooo Apple just trademarked xrOS in their SF Pro font! It’s happening guys! Via @ParkerOrtolani pic.twitter.com/wdx4Q3RbQEMay 16, 2023
See more
While not a typical Apple approach, this is something it has done previously for the Apple TV and other products. A full version of xrOS 1.0 could then follow when the headset hits shelves in late 2023.
The software's name now at least looks set in stone. As spotted by Parker Ortolani on Twitter on May 16, Apple trademarked the 'xrOS' name in its traditional 'SF Pro' typeface in New Zealand, via a shell company.
We'd previously seen reports from Bloomberg that 'xrOS' would be the name for Apple's mixed-reality operating system, but the timing of this discovery (and the font used) bolster the rumors that it'll be revealed at WWDC 2023.
(Image credit: Future)
A report from Apple leaker Mark Gurman on December 1, 2022, suggested that Apple had “recently changed the name of the operating system to “xrOS” from “realityOS,” and that the name stands for “extended reality”. This term covers both augmented reality (which overlays information on the real world) and virtual reality, a more sealed experience that we're familiar with on the likes of the Meta Quest 2.
While xrOS is expected to have an iOS-like familiarity – with apps, widgets and a homescreen – the fact that the Apple AR/VR headset will apparently run both AR and VR experiences, and also use gesture inputs, explains why a new operating system has been created and will likely be previewed for developers at WWDC.
What is xrOS?
Apple's xrOS platform could take advantage of the AR/VR headset's unique hardware, which includes an array of chips, cameras and sensors. It's different from ARKit, the software that lets your iPhone or iPad run AR apps. Apple's xrOS is also expected to lean heavily on the design language seen on the iPhone, in order to help fans feel at home.
According to Bloomberg's Gurman, xrOS “will have many of the same features as an iPhone and iPad but in a 3D environment”. This means we can expect an iOS-like interface, complete with re-arrangeable apps, customizable widgets and a homescreen. Apple is apparently also creating an App Store for the headset.
Stock apps on the AR/VR headset will apparently include Apple's Safari, Photos, Mail, Messages and Calendar apps, plus Apple TV Plus, Apple Music and Podcasts. App developers will also be able to take advantage of its health-tracking potential.
Gurman says that the headset experience will feel familiar to Apple fans – when you put it on, he claims that “the main interface will be nearly identical to that of the iPhone and iPad, featuring a home screen with a grid of icons that can be reorganized”.
But how will you type when wearing the Apple Reality Pro (as it's rumored to be called)? After all, there probably won't be any controllers.
The Sightful Spacetop (above) gives us a glimpse of how the Apple AR/VR headset could work us a virtual Mac display. (Image credit: Sightful)
Instead, you'll apparently be able to type using a keyboard on an iPhone, Mac or iPad. There's also the slightly less appealing prospect of using the Siri voice assistant. Apple is rumored to be creating a system that lets you type in mid-air, but Gurman claims that this feature “is unlikely to be ready for the initial launch”.
It's possible that you'll be able to connect the headset to a Mac, with the headset serving as the Mac's display. We've recently seen a glimpse of how this might work with the Spacetop (above), a laptop that connects to some NReal AR glasses to give you a massive 100-inch virtual display.
What apps will run on xrOS?
We've already mentioned that Apple's AR/VR headset will likely run some optimized versions of existing stock apps, including Safari, Photos, Mail, Messages, Contacts, Reminders, Maps and Calendar.
But given that those apps aren't exactly crying out for a reinvention in AR or VR, they're likely to be sideshows to some of the more exciting offerings from both Apple and third-party developers.
So what might those be? Here are some of the most interesting possibilities, based on the latest rumors and what we've seen on the likes of the Meta Quest Pro.
1. Apple Fitness Plus
Apps like Litesport (above) give us a glimpse of AR fitness experiences that could arrive of Apple’s headset. (Image credit: Litesport)
Assuming the Apple AR/VR headset is light and practical enough for workouts – which is something we can't say for the Apple AirPods Max headphones – then it definitely has some AR fitness potential.
According to a report from Bloomberg's Mark Gurman on April 18, Apple is planning to tap that potential with “a version of its Fitness+ service for the headset, which will let users exercise while watching an instructor in VR”.
Of course, VR fitness experiences are nothing new, and we've certainly enjoyed some of the best Oculus Quest fitness games. An added AR component could make them even more powerful and motivating, with targets added to your real-world view.
2. Apple Freeform
(Image credit: Apple)
We called Apple's Freeform, which gives you a blank canvas to brainstorm ideas with others, “one of its best software releases in years”. And it could be taken to the next level with a version of AR or VR.
Sure enough, Bloomberg's aforementioned report claims that “Apple is developing a version of its Freeform collaboration app for the headset”, which it apparently “sees as a major selling point for the product”.
Okay, work-themed AR/VR work experiences might not sound thrilling and we certainly had misgivings after working for a whole week in VR with the Meta Quest Pro. But mixed-reality whiteboards also sound potentially fun, particularly if we get to play around with them in work time.
3. Apple TV Plus
(Image credit: NextVR)
Because Apple's headset will have a VR flipside to its AR mode, it has huge potential for letting us watch TV and video on giant virtual screens, or in entirely new ways. This means that Apple TV Plus will also likely be pre-installed in xrOS.
Another claim from that Bloomberg report on April 18 was that “one selling point for the headset will be viewing sports in an immersive way”. This makes sense, given Apple already has deals for Major League Baseball and Major League Soccer on Apple TV Plus.
While we haven't been blown away by our experiences with VR meetings in Horizon Workrooms on the Meta Quest, the Apple mixed-reality headset will apparently deliver a next-gen version of FaceTime – and the Reality Pro's hardware could take the whole experience up a notch,
With an earlier report from The Information suggesting that Apple's headset will have at least 12 cameras (possibly 14) to track your eyes, face, hands and body, it should do a decent job of creating a 3D version of you in virtual meeting rooms.
We still haven't really seen a major real-world benefit to VR video meets, even if you can do them from a virtual beach. But we're looking forward to trying it out, while crossing our virtual fingers that it works more consistently than today's non-VR FaceTime.
5. Adobe Substance 3D Modeler
Adobe has already released some compelling demos, plus some beta software called Substance 3D Modeler (above), showing the potential of its creative apps in VR headsets. Will that software's list of compatible headsets soon include the Apple Reality Pro? It certainly seems possible.
The software effectively lets you design 3D objects using virtual clay in a VR playground. Quite how this would work with Apple's headset on xrOS isn't clear, given it's rumored to lack any kind of physical controllers.
These kinds of design tools feel like a shoo-in for Apple's headset, given many of its users are already happy to shell out thousands on high-end Macs and MacBooks to use that kind of software in a 2D environment.
ChatGPT has quickly become one of the most significant tech launches since the original Apple iPhone in 2007. The chatbot is now the fastest-growing consumer app in history, hitting 100 million users in only two months – but it's also a rapidly-changing AI shapeshifter, which can make it confusing and overwhelming.
That's why we've put together this regularly-updated explainer to answer all your burning ChatGPT questions. What exactly can you use it for? What does ChatGPT stand for? And when will it move to the next-gen GPT-4 model? We've answered all of these questions and more below. And no, ChatGPT wasn't willing to comment on all of them either.
In this guide, we'll mainly be covering OpenAI's own ChatGPT model, launched in November 2022. Since then, ChatGPT has sparked an AI arms race, with Microsoft using a form of the chatbot in its new Bing search engine and Microsoft Edge browser. Google has also quickly responded by announcing a chatbot, tentatively described as an “experimental conversational AI service”, called Google Bard.
These will be just the start of the ChatGPT rivals and offshoots, as OpenAI is offering an API (or application programming interface) for developers to build its skills into other programs. In fact, Snapchat has recently announced a chatbot 'called My AI' that runs on the latest version of OpenAI's tech.
For now, though, here are all of the ChatGPT basics explained – along with our thoughts on where the AI chatbot is heading in the near future.
What is ChatGPT?
ChatGPT is an AI chatbot that's built on a family of large language models (LLMs) that are collectively called GPT-3. These models can understand and generate human-like answers to text prompts, because they've been trained on huge amounts of data.
For example, ChatGPT's most recent GPT-3.5 model was trained on 570GB of text data from the internet, which OpenAI says included books, articles, websites, and even social media. Because it's been trained on hundreds of billions of words, ChatGPT can create responses that make it seem like, in its own words, “a friendly and intelligent robot”.
(Image credit: ChatGPT)
This ability to produce human-like, and frequently accurate, responses to a vast range of questions is why ChatGPT became the fastest-growing app of all time, reaching 100 million users in only two months. The fact that it can also generate essays, articles, and poetry has only added to its appeal (and controversy, in areas like education).
But early users have also revealed some of ChatGPT's limitations. OpenAI says that its responses “may be inaccurate, untruthful, and otherwise misleading at times”. OpenAI CEO Sam Altman also admitted in December 2022 that the AI chatbot is “incredibly limited” and that “it's a mistake to be relying on it for anything important right now”. But the world is currently having a ball exploring ChatGPT and, despite the arrival of a paid ChatGPT Plus version, you can still use it for free.
What does ChatGPT stand for?
ChatGPT stands for “Chat Generative Pre-trained Transformer”. Let's take a look at each of those words in turn.
The 'chat' naturally refers to the chatbot front-end that OpenAI has built for its GPT language model. The second and third words show that this model was created using 'generative pre-training', which means it's been trained on huge amounts of text data to predict the next word in a given sequence.
An illustration from Google’s 2017 research paper for the Transformer architecture, which ChatGPT is based on. (Image credit: Google)
Lastly, there's the 'transformer' architecture, the type of neural network ChatGPT is based on. Interestingly, this transformer architecture was actually developed by Google researchers in 2017 and is particularly well-suited to natural language processing tasks, like answering questions or generating text.
ChatGPT was released as a “research preview” on November 30, 2022. A blog post casually introduced the AI chatbot to the world, with OpenAI stating that “we’ve trained a model called ChatGPT which interacts in a conversational way”.
The interface was, as it is now, a simple text box that allowed users to answer follow-up questions. OpenAI said that the dialogue format, which you can now see in the new Bing search engine, allows ChatGPT to “admit its mistakes, challenge incorrect premises, and reject inappropriate requests”.
(Image credit: OpenAI)
ChatGPT is based on a language model from the GPT-3.5 series, which OpenAI says finished its training in early 2022. But OpenAI did also previously release earlier GPT models in limited form – its GPT-2 language model, for example, was announced in February 2019, but the company said it wouldn't release the fully-trained model “due to our concerns about malicious applications of the technology”.
OpenAI also released a larger and more capable model, called GPT-3, in June 2020. But it was the full arrival of ChatGPT in November 2022 that saw the technology burst into the mainstream.
How much does ChatGPT cost?
ChatGPT is still available to use for free, but now also has a paid tier. After growing rumors of a ChatGPT Professional tier, OpenAI said in February that it was introducing a “pilot subscription plan” called ChatGPT Plus in the US. A week later, it made the subscription tier available to the rest of the world.
ChatGPT Plus costs $ 20 p/month (around £17 / AU$ 30) and brings a few benefits over the free tier. It promises to give you full access to ChatGPT even during peak times, which is when you'll otherwise frequently see “ChatGPT is at capacity right now” messages during down times.
(Image credit: OpenAI)
OpenAI says the ChatGPT Plus subscribers also get “faster response times”, which means you should get answers around three times quicker than the free version (although this is no slouch). And the final benefit is “priority access to new features and improvements”, like the experimental 'Turbo' mode that boosts response times even further.
It isn't clear how long OpenAI will keep its free ChatGPT tier, but the current signs are promising. The company says “we love our free users and will continue to offer free access to ChatGPT”. Right now, the subscription is apparently helping to support free access to ChatGPT. Whether that's something that continues long-term is another matter.
How does ChatGPT work?
ChatGPT has been created with one main objective – to predict the next word in a sentence, based on what's typically happened in the gigabytes of text data that it's been trained on.
Once you give ChatGPT a question or prompt, it passes through the AI model and the chatbot produces a response based on the information you've given and how that fits into its vast amount of training data. It's during this training that ChatGPT has learned what word, or sequence of words, typically follows the last one in a given context.
For a long deep dive into this process, we recommend setting aside a few hours to read this blog post from Stephen Wolfram (creator of the Wolfram Alpha search engine), which goes under the bonnet of 'large language models' like ChatGPT to take a peek at their inner workings.
But the short answer? ChatGPT works thanks to a combination of deep learning algorithms, a dash of natural language processing, and a generous dollop of generative pre-training, which all combine to help it produce disarmingly human-like responses to text questions. Even if all it's ultimately been trained to do is fill in the next word, based on its experience of being the world's most voracious reader.
What can you use ChatGPT for?
ChatGPT has been trained on a vast amount of text covering a huge range of subjects, so its possibilities are nearly endless. But in its early days, users have discovered several particularly useful ways to use the AI helper.
Other language-based tasks that ChatGPT enjoys are translations, helping you learn new languages (watch out, Duolingo), generating job descriptions, and creating meal plans. Just tell it the ingredients you have and the number of people you need to serve, and it'll rustle up some impressive ideas.
But ChatGPT is also equally talented at coding and productivity tasks. For the former, its ability to create code from natural speech makes it a powerful ally for both new and experienced coders who either aren't familiar with a particular language or want to troubleshoot existing code. Unfortunately, there is also the potential for it to be misused to create malicious emails and malware.
We're also particularly looking forward to seeing it integrated with some of our favorite cloud software and the best productivity tools. There are several ways that ChatGPT could transform Microsoft Office, and someone has already made a nifty ChatGPT plug-in for Google Slides. Microsoft has also announced that the AI tech will be baked into Skype, where it'll be able to produce meeting summaries or make suggestions based on questions that pop up in your group chat.
Does ChatGPT have an app?
ChatGPT doesn't currently have an official app, but that doesn't mean that you can't use the AI tech on your smartphone. Microsoft released new Bing and Edge apps for Android and iOS that give you access to their new ChatGPT-powered modes – and they even support voice search.
The AI helper has landed on social media, too. Snapchat announced a new ChatGPT sidekick called 'My AI', which is designed to help you with everything from designing dinner recipes to writing haikus. It's based on OpenAI's latest GPT-3.5 model and is an “experimental feature” that's currently restricted to Snapchat Plus subscribers (which costs $ 3.99 / £3.99 / AU$ 5.99 a month).
OpenAI's CEO Sam Altman has confirmed that it's working on a successor to the GPT-3.5 language model used to create ChatGPT, and according to the New York Times this is GPT-4.
Despite the huge number of rumors swirling around GPT-4, there is very little confirmed information describing its potential powers or release date. Some early rumors suggested GPT-4 might even arrive in the first few months of 2023, but more recent quotes from Sam Altman suggest that could be optimistic.
For example, in an interview with StrictlyVC in February the OpenAI CEO said in response to a question about GPT-4 that “in general we are going to release technology much more slowly than people would like”.
He also added that “people are begging to be disappointed and they will be. The hype is just like… We don’t have an actual AGI and that’s sort of what’s expected of us.” That said, rumors from the likes of the New York Times have suggested that Microsoft's new Bing search engine is actually based on a version of GPT-4.
While GPT-4 is unlikely to bring anything as drastic as graphics or visuals to the text-only chatbot, it is expected to improve on the current ChatGPT's already impressive skills in areas like coding. We'll update this article as soon as we hear any more official news on the next-gen ChatGPT technology.
1 844.742.7448 
Email Us
Remote Support 
Contact
