Sebastian Raschka

speaker

1,024 appearances 1 recordings 1 series first heard Feb 2026 last heard 1 Feb

Sebastian Raschka’s voice in public audio — every appearance, attributed to the second.

Trend

recordings per month · last 12 months

Recordings per month over the last 12 months — 1 in all, peaking in Feb 2026 with 1.

Appearances

newest first · ▶ plays the moment

Lex Fridman Podcast · #490 – State of AI in 2026: LLMs, Coding, Scaling Laws, China, Agents, GPUs, AGI · 1 Feb 2026 podcast

Maybe the free tier in future will be something like that.

I do think right now, I mean, it's mostly on the proprietary LLM side, but I think we will see more of that in the open source tooling.

And I think, I mean, it is a huge unlock because then you can really outsource certain tasks from just memorization to actual, you know, like instead of having the LLM memorize what is 23 plus 5, just use a calculator.

So you think that can help solve hallucination?

uh not solve it but reduce it um so still the lm needs to know what um like when to ask for a tool call and the second one is well it doesn't mean the internet is always correct you can do a web search but let's say i asked who won the world cup in let's say 1998 it still needs to find the right website and get the right information so you can still go to the incorrect website and give me incorrect information so i don't think it will fully solve that but it is improving it in that sense and um

So another cool paper earlier this year, I think it was December 31st, so it's not technically 2026, but close.

So like the recursive language model, that's a cool idea to kind of take this even a bit further.

So just to explain, so Nathan, you also mentioned earlier, it's harder to do cool research in academia because of the compute budget.

If I recall correctly, they did everything with GPT-5, so they didn't even use local models.

But the idea is,

let's say you have a long context task, instead of having the LLM solve all of it in like one shot or even like in a chain, you break it down into subtasks.

You have the LLM decide when, like what is a good, let's say, subtask and then recursively call an LLM to solve that.

And I think something like that, also then adding tools and each one, maybe you have like a huge Q and A task, each one goes to the web and gathers information and then you pull it at the end together and stitch it back together.

I think there's going to be a lot of unlock using things like that where you don't necessarily improve the LLM itself.

You improve how the LLM is used and what the LLM can use.

One downside right now with tool use is you have to give the LLM permission to use tools.

And that will take some trust, especially if you want to unlock things like having an LLM answer emails for you.

Not even answer, but just sort them for you or select them for you or something like that.

I don't know if I would today give an LLM access to my emails, right?

I mean, it's like a huge risk.

Showing 681–700 of 1,024 · page 35 of 52