NVIDIA and Alibaba are Plotting

Alibaba and NVIDIA are in cahoots, LinkedIn couldn’t care less about your privacy, and Chipotle has a new toy…

Alibaba Cloud and NVIDIA Take Over Autonomous Vehicles

Alibaba and NVIDIA are collaborating on advanced technology for autonomous driving to enhance the development of AI-powered vehicles. The partnership will combine Alibaba’s cloud infrastructure with NVIDIA’s cutting-edge AI chips and software to create a robust platform for automakers and developers.

Key Takeaways:

• Alibaba Cloud and NVIDIA co-developed a Large Multimodal Model (LMM) solution specifically for automotive applications.

• Alibaba Cloud’s Qwen portfolio of LLMs, including Qwen2-7B and Qwen2-VL, has been integrated with NVIDIA’s Drive AGX Orin platform for autonomous vehicles to enhance AI-driven capabilities in autonomous systems.

So What?

The Drive AGX Orin platform powers intelligent automated-driving systems for leading Chinese EV makers like Li Auto and Xiaomi. This marks the first integration of Alibaba Cloud’s large AI models into NVIDIA’s automotive platform to power next-gen EV fleets.

The partnership is incredibly powerful and somewhat overlooked, especially in U.S. media. It’s not just about dominating China’s EV market; these two giants are truly looking to drive global AI innovation forward.

LinkedIn Just Signed You Up to Train Their AI Models

Okay so LinkedIn did something crazy and borderline unethical last week. The platform opted users into training generative AI models on their data… without asking them.

Shady? Definitely.

The platform updated its terms of service, but of course, I don’t think anyone wants to go back and regularly read that.

LinkedIn writes on a help page that it uses generative AI for purposes like writing assistant features. To opt out of LinkedIn’s default AI data usage, go to this page. Toggle it to “Off” to stop LinkedIn from using your data for AI training.

Google DeepMind has AI Correcting its Own Mistakes

Google DeepMind has introduced a novel approach called Self-Correction via Reinforcement Learning (SCoRe), a method to improve LLMs in handling complex tasks like math problems and coding.

Traditionally, LLMs struggle to self-correct without external input, which poses to be a major challenge in AI development.

Without getting super technical on this approach, the LLM learns a strategy for correcting its own mistakes without getting stuck on minor bumps, in the second stage, it gives itself a little treat after several changes for making more accurate changes (reinforcement learning).

The results of the SCoRe method demonstrate a significant improvement in the self-correction performance of LLMs on different tests:

So What?

This significantly boosts the model’s accuracy and self-correction ability, which will ultimately enable better performance in advanced reasoning tasks. If you’re curious to get deeper into the weeds, check out the full paper here!

“Autocado” is Coming to a Chipotle Near You…

Don’t be alarmed if you see a robot peeling avocados in the back of your local Chipotle… Just kidding this thing is pretty stealthy.

Chipotle recently announced that their “Autocado” robotic assistant is in pilot in a California restaurant.

The fast-casual giant reported it was developing “Autocado” last summer, and now we know it takes 26 seconds for it to cut, core, peel an avocado. The robot slashes guacamole prep time in half from 50 minutes to 25 minutes per batch.

Chipotle built this in partnership with Vebu, a product development company that specializes in automation, robotics and AI solutions within the broader food industry.

AI beyond software

This is yet another case of AI operating at the edge in a robot; yes AI will make software better but look at what it’s doing to real objects.

In Chipotle locations across the U.S., Canada, and Europe this year, the company is expected to use approximately 5.18 million cases of avocados, equivalent to 129.5 million pounds of fruit… I’ll let you do the math on the approximated time savings here..

Stay updated with our favorite highlights, dive in for a full flavor of the coverage!

Elon Musk says Grok 3, the next iteration of xAI’s LLM, is training with insanely more compute than Grok 2.

Anthropic has launched a new retrieval mechanism called contextual retrieval, which combines chunking strategies with re-ranking to significantly improve performance.

Luma AI, a startup founded by former Google, Meta, Adobe and Apple engineers, announced an application programming interface (API) for its Dream Machine video generation model.

Entertainment giants Lionsgate are partnering with artificial intelligence (AI) company Runway to allow a new AI model to be trained on their extensive film and TV archive.

Microsoft and BlackRock, alongside other major industry players, have announced an initiative to raise up to $100 billion to build AI-focused data centers and the necessary energy infrastructure to power them.

Another crazy AI roundup, I have nothing else to add (*drops mic*).

Just enjoy this Gemini fail below:

If you have any suggestions or specific feedback, simply reply to this email or fill out this form. Additionally, if you found this insightful, don't hesitate to engage with us on our socials and forward this over to your friends!

You can find our past newsletter editions here!

This newsletter was brought to you by Digestible AI. Was this forwarded to you? Subscribe here to get your AI fix delivered right to your inbox. 👋