Crazy AI news from week 17
Another week with an astonishing number of crazy stories in AI. I’ve collected the ones I find most exciting here:
By Tim Frank Andersen
Moderna goes all in on Gen AI
This week, a collaboration between OpenAI and the pharmaceutical company Moderna was announced. It involves giving all 3,000 employees full access to Chat GPT Enterprise. Their CEO Stéphane Bancel wants to transform every business process in the company and they are already in full swing.
“Eventually, AI will be able to do more and more science. But the best way to get there is to enhance the productivity of people and accelerate what they can do,”
With 750 different Custom GPTs in use, there are already plenty of successful use cases. One of them is a GPT that undergoes large amounts of research to prepare answers to questions from authorities — a process that used to take weeks, but can now be handled in minutes. Bancel’s ambition is for all employees to use ChatGPT at least 20 times a day and for the app to be front and centre on the phone.
Sam Altman, CEO of OpenAI, says that this is a prime example of how he believes ambitious companies should approach AI and a model for how they see a collaboration with OpenAI working.
Read the full inspiring article from the Wall Street Journal here:
https://www.wsj.com/articles/at-moderna-openais-gpts-are-changing-almost-everything-6ff4c4a5
SenseNova 5 from China beats Chat GPT 4Turbo
It’s what many of us have been waiting for: China has launched their first LLM model that beats GPT4 on virtually all parameters. The company Sensetime announced their latest model this week at their DEVday. The model is trained on 10 Tb tokens and has a context window of 200,000 tokens.
After the launch, SenseTime 商汤科技 shares rose by 30%.
It’s almost impossible to verify their measurements, but the test results seem serious, and it just emphasises the arms race that is currently underway to launch the best LLM model.
You can read more here: https://www.prnewswire.com/apac/news-releases/sensetime-launches-sensenova-5-0-with-comprehensive-updates-and-the-industry-leading-cloud-to-edge-full-stack-large-model-product-matrix-302125415.html
Say hello to the Chinese Astrobot S1
Six months ago, the Chinese government made a clear statement that China will be ready to mass produce humanoids (human-like robots) by 2025. And this week, for the first time, we could see how far they have come. Stardust Enterprises announced their upcoming product Astrobot S1 with an impressive video.
As with everything in China, it’s hard to verify exactly how far along they are or who is behind it, but there’s little doubt that China means business and has no intention of losing this very important technology race. The link between the advanced LLM models and the advanced hardware is what will lead to a real alternative to human physical labour.
See how the Astrobot S1 solves a wide range of domestic tasks in this video: https://www.youtube.com/watch?v=AePEcHIIk9s
Read about China’s Humanoid ambitions here: https://www.therobotreport.com/china-plans-to-mass-produce-humanoids-by-2025/
New AI glasses from Rayban
In 20022, Meta launched their first Rayban AR glasses in collaboration with the world’s largest eyewear company. Since then, they’ve continued to develop features and this week they launched generation 2 with built-in AI and a range of new functionalities that are all voice-controlled:
1. ask the AI to tell it what you’re looking at
2. Translate what you see
3. Live streaming and video calls directly from the glasses
4. Take photos and videos with the glasses
5. Integration with WhatsApp and Messenger
6. Play music directly from Spotify
The glasses are currently selling out faster than they can produce them, but I’m aiming to get my hands on a pair in the next month or so and I’ll be back with a review.
Watch Meta’s demo video here:
https://www.youtube.com/watch?v=EIlkkvRE6Bg
Read about the glasses here:
https://about.fb.com/news/2024/04/new-ray-ban-meta-smart-glasses-styles-and-meta-ai-updates/
Rabbit R1 Launched in New York
Another long-awaited AI gadget that has the ambition to replace our smartphone was launched this week. With sales of over 100,000 units and no idea if it even works, it was exciting to see the reactions. And of course, the reactions are mixed, but actually mostly positive. In Denmark, we are eagerly awaiting when our copy will arrive in the post.
Stay tuned for more info.
In the meantime, read a full and thorough review of the product here: https://shorturl.at/btzC9
Adobe Firefly 3
This week Adobe launched version 3 of their AI extension of Photoshop. The first versions have not been particularly practical, but with 3.0 it really starts to look like something useful. Right now, you can only access it if you download the latest beta version of Photoshop.
You can see how to do that here: https://torybarber.com/firefly-3/
Watch a video demo of Firefly 3 here: https://www.youtube.com/watch?v=GrGup4411QM&t=53s
— And read about all the new features here: https://www.theverge.com/2024/4/23/24138011/adobe-firefly-3-ai-model-photoshop-tools-reference-image
Some extra inspirational material
At the Institute of AI, we are very interested in how companies are successfully implementing AI across their organisation and how GenAI is actually being used.
“If your organisation is currently taking a more cautious approach, be aware that many of your employees are using it anyway. Acknowledging and harnessing this latent curiosity — in a way that aligns with your organisation’s IT policies — must be better than just simply outlawing the practice.”
Harvard Business Review has a very relevant article on the topic: https://hbr.org/2024/03/how-people-are-really-using-genai
If you’d like to learn more about how the Transformer part of GPT models actually works — and in a non-technical way, the Financial Times has created an impressive multimedia article that illustrates it in the best possible way: https://ig.ft.com/generative-ai/