Hacker News: gsjbjt

The Continual Learning Problem

gsjbjt — Tue, 21 Oct 2025 17:14:38 +0000

Article URL: https://jessylin.com/2025/10/20/continual-learning/

Comments URL: https://news.ycombinator.com/item?id=45658435

Points: 4

# Comments: 1

What does it take to build a human-like user simulator?

gsjbjt — Fri, 26 Sep 2025 17:08:58 +0000

Article URL: https://jessylin.com/2025/09/25/user-simulators-2/

Comments URL: https://news.ycombinator.com/item?id=45388748

Points: 6

# Comments: 0

User simulators bridge RL with real-world interaction

gsjbjt — Thu, 10 Jul 2025 20:59:06 +0000

Article URL: https://jessylin.com/2025/07/10/user-simulators-1/

Comments URL: https://news.ycombinator.com/item?id=44525542

Points: 1

# Comments: 0

How will LM agents change the Internet?

gsjbjt — Tue, 12 Nov 2024 16:20:58 +0000

Article URL: https://jessylin.com/2024/11/12/co-adapting-human-interfaces/

Comments URL: https://news.ycombinator.com/item?id=42116859

Points: 1

# Comments: 0

New comment by gsjbjt in "Show HN: Voice bots with 500ms response times"

gsjbjt — Fri, 28 Jun 2024 21:41:39 +0000

That's awesome - can you say anything about what datasets this was trained on? I assume something specifically conversational?

Dynalang: Understanding language via prediction in a multimodal world model

gsjbjt — Tue, 05 Dec 2023 18:24:04 +0000

Article URL: https://dynalang.github.io/

Comments URL: https://news.ycombinator.com/item?id=38534896

Points: 2

# Comments: 0

Dynalang: A multimodal AI agent with a language-world model

gsjbjt — Tue, 08 Aug 2023 23:53:26 +0000

Article URL: https://dynalang.github.io/

Comments URL: https://news.ycombinator.com/item?id=37057038

Points: 2

# Comments: 0

New comment by gsjbjt in "Learning to Model the World with Language"

gsjbjt — Sun, 06 Aug 2023 15:24:31 +0000

Site: https://dynalang.github.io/

Tweet: https://twitter.com/realJessyLin/status/1687515845281624064

Learning to Model the World with Language

gsjbjt — Sun, 06 Aug 2023 15:24:31 +0000

Article URL: https://arxiv.org/abs/2308.01399

Comments URL: https://news.ycombinator.com/item?id=37022940

Points: 1

# Comments: 1

Decision-oriented dialogue: LLM agents to assist with everyday decision-making

gsjbjt — Fri, 02 Jun 2023 15:45:52 +0000

Article URL: https://twitter.com/realJessyLin/status/1664410190719111168

Comments URL: https://news.ycombinator.com/item?id=36166717

Points: 3

# Comments: 1

New comment by gsjbjt in "Do large language models need sensory grounding for meaning and understanding?"

gsjbjt — Mon, 27 Mar 2023 03:54:49 +0000

The point is that LLMs can’t backtrack after deciding on a token. So the probability at least one token along a long generation will lead you down the wrong path does indeed increase as the sequence gets longer (especially since we typically sample from these things), whereas humans can plan their outputs in advance, revise/refine, etc.

Skyplane: Fast bulk data transfers between any cloud

gsjbjt — Wed, 05 Oct 2022 18:12:14 +0000

Article URL: https://skyplane.org/en/latest/index.html

Comments URL: https://news.ycombinator.com/item?id=33099135

Points: 4

# Comments: 0

InCoder: A large language model for infilling and generating code

gsjbjt — Wed, 13 Apr 2022 17:44:42 +0000

Article URL: https://twitter.com/dan_fried/status/1514265047761043456

Comments URL: https://news.ycombinator.com/item?id=31017727

Points: 1

# Comments: 0

New comment by gsjbjt in "Google no longer producing high quality search results in significant categories"

gsjbjt — Mon, 03 Jan 2022 01:21:23 +0000

What are the usual ways to discover discord servers (also through search, or some other means)?

New comment by gsjbjt in "To Understand Language Is to Understand Generalization"

gsjbjt — Sat, 18 Dec 2021 04:54:51 +0000

Nice post! I work on NLP and I think a lot of ideas in this post resonate with what I find exciting about working on the intersection of language + the real world: large text datasets as sources of abundant prior knowledge about the world, structure of language ~ structure of concepts that matter to humans, etc.

I feel like the bottleneck is getting access to paired (language, other modality) data though (if your other modality isn't images). i.e. "bolt on generalization" is an intuitively appealing concept, but then it reduces to the hard problem of "how do I learn to ground language to e.g. my robot action space?" I haven't seen a robotics + language paper that actually grapples with the grounding problem / tries to think about how to scale the data collection process for language-conditioned robotics beyond annotating your own dataset as a proof-of-concept. Unlike language modeling / CLIP-type pretraining, it seems (fundamentally?) more difficult to find natural sources of supervision of (language, action). I'd be curious about your thoughts on this!

> When it comes to combining natural language with robots, the obvious take is to use it as an input-output modality for human-robot interaction. The robot would understand human language inputs and potentially converse with the human. But if you accept that “generalization is language”, then language models have a far bigger role to play than just being the “UX layer for robots”.

You should check out Jacob Andreas's work, if you haven't seen it already - esp. his stuff on learning from latent language (https://arxiv.org/abs/1711.00482).

New comment by gsjbjt in "Is there such a thing as good taste?"

gsjbjt — Mon, 15 Nov 2021 18:12:31 +0000

> My theory is that taste is one quality that separates the academics from the business people. Academia doesn't necessitate a lot of taste. If you have it, great. If you don't have it, no big deal.

This might be true for academics in ancient Greek literature, but certainly isn't true for academics in CS nowadays. If you don't have good taste in research problems that are {important for downstream industry applications, scientifically interesting, tractable}, you won't get anything done, and you won't get published. If anything, the pressure for academics to develop good taste is stronger than for people designing product. You can have a product that provides just one utility that users desperately need and have terrible taste for all the other axes that make a product "good," and do just fine. Academic papers get judged (in peer review / traction after publication) purely against the taste and aesthetics of other people in your community.