Embedding models are one of the most important parts of model development and deployment. You have to encode all your data in the first place. But while GPT-4 is well ahead of open source competition in LLMs, ada-002 didn't necessarily have that level of a commanding lead.
@mattlynley
-
OpenAI Cuts Ada Embedding Model Price by 75%
By
–
Today's issue is about what seems to be a grand tradition in AI announcements now. OpenAI, talking about a slew of product updates, said in a throwaway line it was cutting the price of its embedding model ada-002 by a whopping 75%.
-
Training Data Impact on Gemini from User Behavior Changes
By
–
I suspect this is absolutely the least of their worries with this. If I were there I'd be more wondering about the permanent behavior change this might create on the site and if that affects data that could be used in gemini (if the training set isn't already done).
-
Google’s Corporate Development Strategy for Reddit Acquisition
By
–
I think the crowd that is like "Google should buy Reddit because search results" is still underestimating how good Google's corp dev team is. This is the same group that got Google to be the default engine in iOS. They will get a preferred API in some form.
-
Oracle Earnings: Key Takeaway on AI Strategy
By
–
Doing the whole "this is really all you need care about" thing again for the Oracle earnings this time today with this line right here:
-
GPU Shortage Forces Industry-Wide Changes in AI
By
–
You can ready the full (paywalled) story here: https://
supervised.news/p/how-the-gpu-
crunch-is-forcing-the
… -
Nvidia dominates AI hardware market while AMD struggles to compete
By
–
Nvidia has spent years building up its AI hardware business and equipment, and it now holds a dominant position in that industry. It's also let AMD co-exist for years, which has been developing quickly but faces an enormous uphill battle to unseat Nvidia.
-
Nvidia Seeds CoreWeave with GPUs to Build Competitive Ecosystem
By
–
In addition, I've heard Nvidia is seeding CoreWeave (and potentially others) with GPUs to build competition in the ecosystem as the largest cloud providers scramble for access to that hardware. Nvidia is a big investor in CoreWeave.
-
Developers Shift to CPU-Based Inference and Model Fine-Tuning
By
–
That's forcing developers to get creative, including shifting their focus to fine-tuning models—and moving more work over to CPUs. As one source told me about increasing CPU usage: "would you prefer slow inference, or no inference?"
-
GPU Shortage Impacts AI Development at Weights & Biases Conference
By
–
Today's issue is about the growing GPU crunch, which was on a lot of minds at the Weights & Biases conference this week. Nvidia hardware—preferred for PyTorch development—is increasingly difficult to get a hold of, creating a more challenging AI development environment.