Welcome to a roundup of this week’s juiciest AI news.
This week AI starred in Google’s fake video.
Reformed music pirates get back in the saddle.
And ChatGPT may have become a little sad and lazy.
Let’s dig in.
Is it cheating?
Oxford University now allows students to use AI for its Economics and Management course. Students were probably already doing this but now they’ve been given a license to do so openly. There are some interesting reasons why AI could make it more difficult to turn in a decent paper though.
Is it cheating when students use AI? Where’s the line? A new study reveals that there’s a lot of confusion surrounding generative AI in education.
As ChatGPT handles more of the heavy lifting, students may lose some basic skills. But does it matter? We may have lost the knack of starting a fire by banging rocks together because now we have matches and lighters. I’m ok with that.
Lies or just marketing?
Google copped a fair amount of criticism for its misleading Gemini marketing video when it announced its long-awaited new models. There was more than a little sleight of hand involved in wowing us with Gemini’s performance.
When you see the real interactions Google’s engineers had with Gemini it makes the video a lot less impressive.
Despite the drama, Gemini does seem to be very good. Google has integrated its Gemini models into coding and development tools. If you’re a software developer then Duet AI for Developers is about to make your job a lot easier.
LimeWire back in the music game
We’re not saying you did but, hypothetically, if you were the kind of person who pirated music in the early 2000s, you probably used Napster or LimeWire to download MP3s. After a decade of absence, LimeWire is making a comeback with its AI-generated music platform.
Meta made a few tweaks to its AI Experiences including releasing its image generator as a standalone platform. Imagine with Meta AI creates great pictures and your holiday snaps may have had something to do with that. You still can’t copyright AI-generated images in the US, but other countries may take a different view.
China doesn’t have the best reputation for enforcing copyright law so we were a little surprised when a Beijing court ruled that AI art is protected by copyright. The incident that led to the ruling seems trivial but the ramifications could be huge.
Hey AI, are you ok?
ChatGPT seems to have been calling it in lately. Users have said the chatbot has become lazy since the beginning of December. Could ChatGPT be on a winter break and feeling the effects of SAD like humans do?
We know xAI’s Grok chatbot often responds in, shall we say, unconventional ways. But this week things got a little awkward. Grok appeared to copy-paste from OpenAI’s policy, roasted Elon Musk, and seemed to be leaning to the political left. Grok, are you ok?
AI on the job
Like ChatGPT, we may feel a little lazy this time of year but we’d still like to keep our jobs if AI will let us. Microsoft and the biggest US labor group announced a partnership to address how AI impacts workers’ rights as anxiety about AI replacing workers grows.
OpenAI employees did a little collective bargaining of their own during the Sam Altman ‘You’re fired. You’re hired.’ drama a few weeks back. It turns out that there was more to the open letter than they let on.
Mistral in the mix
An AI startup that few had heard of is now grabbing headlines. The rise of the French AI startup, Mistral AI has been meteoric. The open-source startup secured $415M in funding taking it from zero to a $2 billion valuation in 6 months.
Mistral’s new sparse Mixture of Experts LLM beats GPT-3.5 in a series of benchmark tests. If you want to understand the secret sauce in Mixtral 8x7B then check out our crash course on Mixture of Experts and Sparsity. You’ll sound really smart when you explain this to your colleagues.
Brainoware – It’s alive!
There’s no need to fret over Nvidia’s long lead times when you can run your AI in human brain cells in a petri dish. These researchers dubbed their mini-brain “Brainoware” and taught it to perform AI speech recognition.
A new study found that AI matches doctors in X-ray analysis. The AI system analyzes an X-ray image and then scans for 37 different conditions the patient may have. The accuracy it achieves compared to human doctors is really impressive.
AI in the middle
The printing press was banned in the Middle East for 200 years but countries like the UAE are making up for lost time by accelerating their homegrown AI industry. The US is getting a little twitchy over China’s involvement in the region though.
Global AI regulation may be a long way off but, against all odds, the EU’s landmark agreement on AI was finally passed. The EU AI Act will provide an interesting backdrop to the upcoming European Chatbot & Conversational AI Summit 2024 that will be held in March.
In other news…
Here are some other clickworthy AI stories we enjoyed this week:
- McDonald’s is partnering with Google to deploy generative AI in its stores to bring you “hotter, fresher food”.
- Researchers created a diffusion model that upscales the resolution of videos.
🔬 Our study introduces “Upscale-A-Video,” a text-guided latent diffusion framework for video upscaling. It ensures temporal coherence locally & globally, balancing fidelity and quality.
🚀 Project page: https://t.co/3UkaXXyMCC
💻 GitHub: https://t.co/irQRuPHxED
🎥 Video:… pic.twitter.com/A7Ymz3be3Q— Chen Change Loy (@ccloy) December 12, 2023
- W.A.L.T is a transformer-based approach for photorealistic text-to-video generation. This is very cool.
- Sports Illustrated’s CEO has been fired after the AI scandal at the media company. We covered the AI journalistic snafu that led to this here.
- What does an AI model look like? This LLM visualization is amazing!
- Is Elon Musk looking to poach Ilya Sutskever from OpenAI?
And that’s a wrap.
In a world where AI fakes proliferate, we’re still feeling a little bitter about Google’s Gemini video. Let’s hope that Gemini Ultra is worth the wait when it’s finally released.
Have you noticed ChatGPT being a little sluggish, or is it just us? Maybe it’s just a little nervous about being knocked off the top of the AI ladder by GPT-5 pretty soon.
If you generated a decent track on LimeWire please send us the link along with any leads for other AI news we should be covering.
