Underrated decision by OpenAI that kicked off the LLM explosion: having GPT-3.5 accessible through a chatbot, rather than the API playground. Computer scientists tend to think in APIs, but they make access to tools hard for non-coders, who don't think that way. Chat was smart.
@emollick
-
Copilot Variants: Enterprise, Pro, and Windows Editions
By
–
It might also mean the general concept of Copilot. Or the Office Copilot, Enterprise Edition, or Copilot Pro. Also Copilot for windows.
-
Frontier models outperform specialized AIs trained on proprietary data
By
–
There was a moment that we thought proprietary data would let organizations train specialized AIs that could compete with frontier models. It turns out that probably isn't going to happen. The largest frontier models are just much better at most complex tasks than smaller models.
-
GPT-4 Outperforms Specialized Finance AI Model
By
–
This remains one of the most consequential experiments in AI: Bloomberg trained a GPT-3.5 class AI on their own financial data last year… …only to find that GPT-4 8k, without specialized finance training, beat it on almost all finance tasks. Hard to beat the frontier models.
-
Enterprise demand outpaces consumer adoption in LLM market
By
–
And the other factor was that they built for a consumer market, making a big bet (which seemed sensible) that this would go first. Instead, enterprise turned out to be the place where the big demand was Pi was a very cool product, but I suspect the other LLMs can do similar chat
-
Inflection’s Challenge: Competing in the GPT-4 Class Model Race
By
–
The thing was Inflection shot its shot. Their latest model was very good, but wasn't GPT-4 class. Given that there are now three GPT-4 class models out there (with Lllama-3 coming soon) and two GPT-4.5/5 (?) class models coming, I wonder if they could have caught up at this stage
-
VC discusses AI future trajectory in upcoming conversation
By
–
Really looking forward to joining James for this conversation. He is one of the rare venture capitalists who has a strong sense of where the AI world is heading, and it should be an interesting discussion.
-
Enterprise AI Solutions: Beyond Basic LLM Integration and RAG
By
–
Everyone makes fun of startups that are just "GPT-4 wrappers" but the bigger issue are large enterprise software companies slapping chatbots powered by GPT-3.5 class LLMs with basic RAG into their products and claiming they have an AI solution At least the startups are ambitious
-
Workers Secretly Benefiting from Generative AI Automation Now
By
–
The surplus value created by generative AI goes to workers a surprising amount at this point. I talk to people all the time who have automated parts of their job and aren’t telling management. Eventually companies will catch up but a lot of people are secretly benefiting now.
-
Compute Moat: Key Advantage for AI Labs Until Scaling Falters
By
–
Until the scaling law for AI training starts to falter, compute is the moat and source of advantage in AI. And two of the top AI labs have no compute of their own.