latest
Outsourcing to third-parties is the best way for Apple to move forward with AI
Apple's track record with AI and ML in Siri doesn't mode well for the future. Outsourcing to a company like Google or OpenAI is a far better solution.
Artificial intelligence is part of the conversation when talking about virtually all parts of the tech industry in 2024. Despite that being the case, Apple has been relatively quiet on the AI/ML front. This is all while competitors like OpenAI, Google, and Microsoft are racing to ship AI products. In fact, although the measure did not pass, 37.5% of Apple shareholders voted in favor of conducting a report concerning the company's AI efforts. People are concerned that Apple may be asleep at the wheel when it comes to AI, but that isn't exactly the case. Apple CEO Tim Cook has been steadfast in his claim that the company will ship AI features on the iPhone 16 series later this year. To do so, Apple might follow an old adage: if you can't beat 'em, join 'em.
5 things we want to see from GPT-5
GPT-5 is expected to arrive this summer, and these are the big things we want to see it arrive with
GPT-5 has been a hot topic for quite a while now, and OpenAI CEO Sam Altman recently made comments regarding the future of the GPT model on Lex Fridman's podcast. In that podcast, he stated that GPT-4 "kind of sucks" now and that he's looking forward to what comes next. He refused to refer to it as "GPT-5", but a recent report from Business Insider did name it as such, with people familiar with the LLM referring to it as "materially better" when compared to GPT-4.
Claude 3 Opus vs Microsoft Copilot Pro: One clear winner
Claude 3 Opus is a new LLM that's making waves, but how does it fare against Microsoft's Copilot Pro subscription?
The world of large language models (LLMs) has been an oligopoly of essentially three major players: OpenAI, Microsoft, and Google. Now it's growing with a new LLM from Anthropic, dubbed Claude 3 Opus. It's an LLM that's been making waves with answers that many claim are better than GPT-4. GPT-4 Turbo is what powers Microsoft Copilot, and how does Microsoft Copilot Pro stack up against Claude 3 Opus? There's a clear winner.
Claude 3 Opus vs Gemini Advanced: A surprisingly close call
Claude 3 Opus is a groundbreaking new AI model that manages to really impress.
In the world of large language models, it's been dominated by ChatGPT, Microsoft Copilot, and Google Gemini. If you've heard of Anthropic, its new LLM dubbed "Claude" has been doing rounds as a potential major competitor in the space. Claude 3 Opus recently launched, and it's managed to really, really impress. This is how it stacks up against Google Gemini Advanced in particular.
Apple's MM1 language model could finally make Siri useful
Apple's MM1 language model could be developed in part to make Siri useful as part of the iPhone 16 series.
Apple has been quietly working on generative AI for quite a while now, with company CEO Tim Cook making the announcement in a recent earnings call. Nobody really knows what that means just yet, but a recent research paper published without much fanfare by engineers at the company can give us some clues. Apple's MM1 language model is a multimodal model with up to 30 billion parameters, and can also come packaged as a mixture of experts model (MoE) that goes up to 64 billion parameters.
GPT-5 might finally launch this summer
GPT-5 is the next big update to GPT from OpenAI, and it might be just a few months away
GPT-4 is still one of the best large language models out there, and it powers Microsoft Copilot and ChatGPT Plus. However, according to Business Insider, we may see GPT-5 arrive as soon as this summer. One source for the site stated that GPT-5 is "materially better," with the AI model being demonstrated in use for data and utility specific to his company. Given the growing advancement from competitors like the Gemini Ultra model and Claude 3 Opus, OpenAI is likely starting to feel the mounting pressure.
Generative AI's "model collapse" will cause it to poison itself, here's what that means
Model collapse is an impending issue when it comes to AI, and here's what it means.
Generative AI has taken off in ways that we could have never imagined, and it's become incredibly pervasive across the entire internet. With X (formerly Twitter) bots powered by generative AI, automated accounts on Reddit, and even scholarly papers being made at least in part by LLMs, the internet is slowly being flooded with content generated by algorithms. While humans can by and large filter this content out as they go (or even find some value in it), LLMs can't, and it's exactly why AI will poison itself over time.
Claude 3 Opus vs ChatGPT Plus: A challenger approaches
Claude 3 Opus and ChatGPT Plus are two subscription model LLMs, but the differences between them are quite stark.
In the world of large language models, it's been a realm purely dominated by ChatGPT, Microsoft Copilot, and Google Gemini. There's a new player in town from Anthropic though, and it's dubbed "Claude." Claude 3 Opus recently launched, and it's giving a run for its money over competitors in a way that seemingly puts it ahead of even the best of what GPT-4 has to offer.
AI can hallucinate, and here's one way I experienced it
LLMs are widely known to hallucinate, and here's a fun and interesting way I experienced it that exposes some gaps in AI knowledge.
When it comes to AI, there's a lot about it that we don't really know. Reverse engineering AI models is a massive industry, and the field itself is called mechanistic interpretability. With that said, we have some understandings on a conceptual level of how some things happen, and an interesting one I came across recently was a strange hallucination from OpenAI's GPT-3.5 when I asked it to translate a word into Irish.
Claude 3 Opus is an LLM that outperforms GPT-4 and Gemini
If you've been hearing about Claude 3 Opus recently but aren't sure what it is, we have a basic hands-on to show you what all the hubbub is about!
If you've been following the generative AI space for a while, then you may have heard of Claude 3 Opus. Seemingly coming out of nowhere, it's an LLM by Anthropic that manages to compete with GPT-4, Google Gemini, and Microsoft Copilot in a meaningful way. Not only that, but many people feel confident in saying that it outperforms GPT-4, and to be honest, I absolutely agree. Claude 3 Opus is the real deal, and it's a surprisingly powerful model that manages to outperform GPT-4 in a ton of ways.
Apple will launch generative AI features later this year, but nobody knows what that means
Apple has been known to march to the beat of its own drum, but for once, it's starting on a backfoot.
Generative AI is all the rage, and companies across the world are all trying to incorporate new features that make use of it. While Apple has typically marched to the beat of its own drum, it turns out that not even the Cupertino giant is exempt. Tim Cook, in a recent quarterly earnings call, said that Apple's own generative AI will come "later this year", but what that means is anyone's guess.
OpenAI's impressive text-to-video generator Sora is launching publicly this year
OpenAI's CTO has confirmed that Sora is launching this year, and that the company is working on adding new capabilities to the model too.
Last month, ChatGPT's creator, OpenAI, once again shocked the world with its latest AI model; a text-to-video generator called Sora. A demo from the company showed how Sora could generate almost flawless videos and simulations of moving objects - such as a woman walking in Tokyo - through just text prompts.
Meta's LlaMa 3 is expected this year, and these are 5 things we'd like to see
Meta's LlaMa 3 is expected to arrive as early as June, and there's a lot that we want to see from it.
The artificial intelligence LLM race has been slowly heating up since its jumpstart with ChatGPT. We've seen Google's Bard become Gemini, Microsoft's Bing Chat become Copilot, and we've seen Meta release its open source Llama 2 model that anyone can run on their own computer. Now, as the competition advances, Meta is expected to launch Llama 3 later this year — possibly as soon as July according to reports.
Does AI risk consumer burnout?
AI is clearly the future, but bringing affordable products to market is taking time.
It's been over a year since ChatGPT was released in November 2022, kicking off a wave of hype around AI technologies to rival the early internet. But today, for all the hype, investment, and discussion around AI, the world hasn't quite changed yet. Many of the demos we've seen of the capabilities of AI for consumers have been scaled back significantly, and many early entries into the new market of AI products have struggled to gain traction.
ChatGPT and Gemini are cool, but they're not where the future of AI is heading
The future of AI is almost certainly going to be models trained for specific tasks, rather than general-purpose LLMs.
Large Language Models (LLMs) are powerful tools, and ChatGPT, Microsoft Copilot, and Google Gemini consistently manage to blow me away. Their capabilities are extensive, but they're not without their faults. Hallucinations are a big problem with LLMs like these, though companies are aware of them and try to stamp them out whereveer possible. However, I don't think these models are the future of LLMs. I think the future of AI are the smaller, specialized models, rather than the general purpose tools that these are.
Google's Gemini 1.5 Ultra will need to compete with GPT-5, not GPT-4
Google's Gemini 1.5 Pro model wowed us, and the Ultra model could be even better.
There's an ongoing arms race in the artificial intelligence sphere, particularly between big players like OpenAI, Microsoft, and Google. These three companies are locked into a never-ending battle of advancement, and Google is especially an outsider given that it doesn't use OpenAI's GPT technology at all. Google instead has been using its own home-grown technology, initially starting with PaLM and Bard and now more recently switching to Gemini 1.0 and Gemini 1.5. With Gemini 1.5 Pro already seemingly doing the impossible, we've never been more excited for Gemini 1.5 Ultra.
4 weird AI integrations we never thought we'd see
Some AI integration we never thought would be possible a few years ago.
We're witnessing the dawn of AI technology right now, and it's being integrated into various parts of our lives in ways we couldn't have imagined. It's true that AI was being used in some capacity in things like self-driving cars, voice assistants, and chatbots, but it wasn't as prevalent as today. It's one thing to have dedicated services like AI image generators. Still, it's completely different to have AI integrated directly into the products and services we use every day, like web browsers, many of which you may never have expected to become a reality.
Nvidia's new coding LLM will make you a better programmer and can run on a CPU
Nvidia, in collaboration with HuggingFace and ServiceNow, has released StarCoder2, and it will help you generate code.
If you're a programmer looking to make your life a little bit easier, ServiceNow, Hugging Face, and Nvidia have released StarCoder2, an LLM that you can run locally to generate code. Trained on 619 languages, the 15 billion parameter version of StarCoder2 was built by Nvidia and is the largest of the lot. ServiceNow built the three billion parameter version, and Hugging Face built the seven billion parameter version.
Sora is the first true text-to-video model, but it won't be the last
Sora is the first and won't be the last, especially if history is anything to go off of.
Sora was recently unveiled by OpenAI as a true text-to-video generation model, and with it came a number of questions regarding the societal impact such a tool may have. The ability to generate authentic-looking videos with ease raises questions regarding the integrity of video as a form of proof, while also raising questions about where OpenAI trained its model in the first place.
Ambient Computing: What is it?
Ambient computing is underlying a lot of our technology today, but what is it?
Devices all around us are getting smarter, and part of that perceived intelligence is the fact that so many of them are practically autonomous. If you have a Google Pixel, for example, you'll be familiar with features like Now Playing that just run in the background without any input. It's essentially invisible tech.