I'd add JAX to the list as well. In my view all three frameworks are here to stay for the long term and you should have at least some familiarity with all three. We're in a multi-framework world now. For applied ML & production, TF remains the most well-rounded option.
@fchollet
-
Deep Learning Expertise Shortage More Critical Than GPU Scarcity
By
–
The shortage of folks who actually understand how deep learning works and what it can/cannot do is gearing up to be even more dire than the upcoming GPU shortage.
-
LLMs Used for Creating Fake Social Media Accounts and Spam Content
By
–
One of the major applications of LLMs: generating "content" for fake humans on Twitter and other social media platforms that have weak spambot filters — as seen in this thread…
-
Generative AI as a Content Blender Without Personality
By
–
Generative AI feels a bit like the Borg — a big content blender that can assimilate anything and spit back infinite remixes, with no personality of its own.
-
Critique du discours optimiste sur les robotaxis autonomes
By
–
"one million robotaxis" vibe. Magical thinking.
-
Efficient LLM Fine-tuning with LoRA Tutorial on Keras
By
–
Awesome new tutorial on http://
keras.io: how to use LoRA to perform very efficient fine-tuning of LLMs. -
Realistic AI Views: Opportunity Between Hype and Skepticism
By
–
Anyhow — there is still a world of opportunity ahead for those with realistic views about the potential of AI. Believing in magic and fairies is in every way as counter productive as being a tech reactionary.
-
Generative AI: Realistic Understanding Emerging, Hype Expected to Decline
By
–
The good news is that folks are now starting to having a slightly more grounded understanding of what generative can and cannot do. Things are still very bubbly though, and sentiment will come down much further through the rest of the year and next year.
-
AI Investment Expectations Unrealistic, Revenue Gap Massive
By
–
That's because the expectations that people had for AI in Q1-Q2 2023 were entirely unmoored from reality. Investor expectations especially. It will take many years for AI-driven revenue to recoup the investment poured into AI in 2022-2023. Right now they differ by 10-20x.
-
AI adoption is a multi-decade trend despite current bubble deflation
By
–
The adoption of AI is a multi-decade trend that is only just getting started. The scope of AI use cases, the amount of revenue generated by AI apps will increase maybe ~10x in the next 5-10 years. And yet, AI is a bubble — which is now starting to slightly deflate.