fsndz

13 ms·

121.
by fsndz 2y ago
isn't that what declarative programming frameworks do already ?
122.
by fsndz 2y ago
They want to throw LLMs at everything even if it does not make sense. Same is true for all the AI agent craze: https://medium.com/thoughts-on-machine-learning/langchains-s...
123.
by fsndz 2y ago
Did the same and gpt-3.5-turbo-instruct still lost all the games. maybe a diff in stockfish version ? I am using stockfish 16
124.
by fsndz 2y ago
I know right ?
125.
by fsndz 2y ago
Sam Altman might be wrong then? Learning from data is not enough; there is a need for the kind of system-two thinking we humans develop as we grow. It is difficult to see how deep learning and backpropagation alone will help us model that.
126.
by fsndz 2y ago
Did the same and gpt-3.5-turbo-instruct still lost all the games. maybe a diff in stockfish version ? I am using stockfish 16
127.
by fsndz 2y ago
PS: I ran and as suspected got-3.5-turbo-instruct does not beat stockfish, it is not even close "Final Results: gpt-3.5-turbo-instruct: Wins=0, Losses=6, Draws=0, Rating=1500.00 stockfish: Wins=6, Losses=0, Draws=0, Rating=1500.00&quot
128.
by fsndz 2y ago
PS: I ran and as suspected got-3.5-turbo-instruct does not beat stockfish, it is not even close "Final Results: gpt-3.5-turbo-instruct: Wins=0, Losses=6, Draws=0, Rating=1500.00 stockfish: Wins=6, Losses=0, Draws=0, Rating=1500.00&quot
129.
by fsndz 2y ago
wow I actually did something similar recently and no LLM could win and the centipawn loss was always going through the roof (sort of). I created a leaderboard based on it. https://www.lycee.ai/blog/what-happens-when-llm
130.
by fsndz 2y ago
damn you really don't like Hegel, do you ?
131.
by fsndz 2y ago
exactly or one can say that all artificial intelligences are and will be human intelligence amplifiers
132.
by fsndz 2y ago
absolutely, directionally it is useful.
133.
by fsndz 2y ago
I came to the same conclusion a while back. LLMs are very useful when user expertise level is medium to high, and task complexity is low to medium. Why ? because it those scenarios, the user can use the LLM as a tool for brainstorming on dr
134.
by fsndz 2y ago
that's a nice strategy for a 39 person startup ! using the tools is indeed the only way to see what works and what doesn't.
135.
by fsndz 2y ago
I am not surprised at all. The excitement is receding in accordance with the hype cycle theory. People tried the tools and saw their worth and the limitations. This is actually good news, it means we are entering a moment of truth, a moment
136.
by fsndz 2y ago
so for you it's bullshit ? or a kind of rationalisation a posteriori ?
137.
Anyone who believes in the Hegelian concept of "cunning of reason"?
4 points by fsndz 2y ago|9 comments
138.
by fsndz 2y ago
nice. a similar idea was recently used to detect ragallucinations. the key is using logits when provided It was super insightful reading the clash eval paper https://www.lycee.ai/blog/rag-ragallucinations-and-how-to-fi.
139.
by fsndz 2y ago
this is not AI slop. do you have a proof of what you just wrote ?
140.
by fsndz 2y ago
More AI sauce won't hurt right ? Meanwhile, we still have to solve the ragallucination problem. https://www.lycee.ai/blog/rag-ragallucinations-and-how-to-fi...
141.
by fsndz 2y ago
Semantic search is a powerful tool that can greatly improve the relevance and quality of search results by understanding the intent and contextual meaning of search terms. However, it’s not without its limitations. One of the key challenges
142.
by fsndz 2y ago
tragic
143.
by fsndz 2y ago
I had the exact same experience. Just calling an API call "agent" and giving it a system prompt like "you are a robust data analyst" doesn’t magically make it one. That’s not how it works. My first conclusion is that it’
144.
Crewai Raises $18M–But Are AI Agents Ready for Prime Time?(finance.yahoo.com)
4 points by fsndz 2y ago|3 comments
145.
by fsndz 2y ago
Saw that Crewai just raised $18 million for their multi-agent platform, but I’m struggling to understand the value beyond what you can do with vanilla API calls to OpenAI or something like DSPy/OpenAI Swarm. From what I’ve seen, AI age
146.
by fsndz 2y ago
> My personal take, we’ve made the cost of failure to high and cheating too easy. This is so true. I was recently pondering about the impact of AI cheating in Africa and came up with the conclusion that it won't be as significant as
147.
by fsndz 2y ago
Building robust LLM-based applications is token-intensive. You often have to plan for the parsing and digestion of a lot of tokens for summarization or even retrieval augmented generation. Even the mere generation of marketing blogposts con
148.
by fsndz 2y ago
Yes. Company continue to pay them for political reasons mostly. Just look at how many McKinsey partners became CEOs and failed terribly at it (latest example Starbucks
149.
by fsndz 2y ago
absolutely !
150.
by fsndz 2y ago
more proof that we need more than LLMs to build LRMs: https://www.lycee.ai/blog/drop-o1-preview-try-this-alternati...

More ›