jsemrau

14 ms·

181.
by jsemrau 2y ago
Thanks. I was rushing out to the gym.
182.
by jsemrau 2y ago
Agents are still a misaligned concept in AI. While this article offers a lot in orchestration, memory (only mentioned once in the post) and governance are not really mentioned. The latter is important to increase reliability -- something Il
183.
by jsemrau 2y ago
If it could replace a human employee, what would the right price tag be?
184.
by jsemrau 2y ago
MySQL has this as default as far as I recall. But then I never delete, I just set "deleted" to yes.
185.
by jsemrau 2y ago
Game Theory and Agent Reasoning?
186.
by jsemrau 2y ago
90% of Foursquares revenue comes from Enterprise clients. This dataset would not cannibalize that revenue, but it would provide the general population by fixing bugs and finding new use-cases that might put them in a better spot when compe
187.
by jsemrau 2y ago
should have taken a screenshot.
188.
by jsemrau 2y ago
The Harry Potter game had much better AR integration
189.
by jsemrau 2y ago
Google starts getting into the way too much. Recently I searched for German football clubs that "fan rivalries" and Google refused that and only gave me results for "fan friendships" .
190.
by jsemrau 2y ago
I'd argue longform content has done more for opinion-making than microblogs in the 2024 election. And to be honest, I like that. Irrespective of the politics the podcasts with JD Vance and even Trump at Theo Von were remarkable. Seeing
191.
by jsemrau 2y ago
There are many ways to test for reasoning and deterministic computation as my own work in this space has shown .
192.
by jsemrau 2y ago
Please, someone think of the Math reasoners.
193.
by jsemrau 2y ago
That might make sense.
194.
by jsemrau 2y ago
To my understanding yes. But I never found a good use-case for sentiment classification.
195.
by jsemrau 2y ago
One of the main reasons I am bearish on Sakana.AI is that for a Japanese-first LLM to work you need to solve a context sensitive language where all 4 alphabets intermingle in a given body of text. So when you are hiring Japanese-speaking Ph
196.
by jsemrau 2y ago
When I worked at Electronic Arts in 2003 we had a lovely restaurant on the premises.
197.
by jsemrau 2y ago
You are probably right. I approached it from a high-value dataset perspective but would agree that fuzzy frames probably don't help much.
198.
by jsemrau 2y ago
Makes you wonder if its possible with the right seed, scheduler, and prompt to complete recreate the original.
199.
by jsemrau 2y ago
Maybe its a combination of both.
200.
by jsemrau 2y ago
Ah, this is where this comes from. There has been rumours flying around in Stable Diffusion / Flux circles that you would get much more realistic pictures when you include a photo id like IMG_0416.
201.
by jsemrau 2y ago
I have lived in Japan for many years. There is a certain phenomena where foreigners when they meet each other in the supermarket experience a moment of awkwardness like we entered each others TikTok feed. You don't know if one should s
202.
by jsemrau 2y ago
On narrow AI or generative AI Risk Management?
203.
by jsemrau 2y ago
I loved they keyboard of my Blackberry tbh. Why can't a future iPhone can have a slidable keyboard?
204.
by jsemrau 2y ago
I think it can be quite interesting especially if you consider different character types (in Anthropic lingo this "personality"). The only problem right now is that using a proprietary LLM is incredibly expensive. Therefore having
205.
by jsemrau 2y ago
Memory is really interesting. For example, if you play 100,000 rounds of 5x5 Tic Tac Toe. Do you really need to remember game 51247 or do you recognize and remember a winning pattern? In Reinforcement Learning you would based on each win re
206.
by jsemrau 2y ago
>If only I could get management to understand that a bunch of prompts shitting into eachother isn't "cutting-edge agentic AI"... It should never be this way. Even with narrow AI, there needs to be a governance framework th
207.
by jsemrau 2y ago
You are on the right track in my opinion. The key is to encode the interface between the game and the agent so that the agent can make a straightforward choice. For example, by giving the agent the state of a nxn board as the world model, a
208.
by jsemrau 2y ago
I am playing around with LLMs and Game Theory. Currently, I let them play 5x5 board games in a tournament against Reinforcement Learning Agents and a Random Player. I am capturing the "thoughts" of the agent for behavioral data an
209.
by jsemrau 2y ago
>I'm sure they are looking to LLM proposals Well, it is a use case for this challenge https://www.kaggle.com/competitions/gemini-long-context
210.
by jsemrau 2y ago
Claude has personality. I think that was one of the more interesting approaches from them that went into my own research as well.

More ›