Sveriges mest populära poddar

Machine Learning Street Talk (MLST)

#80 AIDAN GOMEZ [CEO Cohere] - Language as Software

52 min • 15 november 2022

We had a conversation with Aidan Gomez, the CEO of language-based AI platform Cohere. Cohere is a startup which uses artificial intelligence to help users build the next generation of language-based applications. It's headquartered in Toronto. The company has raised $175 million in funding so far.

Language may well become a key new substrate for software building, both in its representation and how we build the software. It may democratise software building so that more people can build software, and we can build new types of software. Aidan and I discuss this in detail in this episode of MLST.

Check out Cohere -- https://dashboard.cohere.ai/welcome/register?utm_source=influencer&utm_medium=social&utm_campaign=mlst

Support us!

https://www.patreon.com/mlst 

YT version: https://youtu.be/ooBt_di8DLs

TOC:

[00:00:00] Aidan Gomez intro

[00:02:12] What's it like being a CEO?

[00:02:52] Transformers

[00:09:33] Deepmind Chomsky Hierarchy

[00:14:58] Cohere roadmap

[00:18:18] Friction using LLMs for startups

[00:25:31] How different from OpenAI / GPT-3

[00:29:31] Engineering questions on Cohere

[00:35:13] Francois Chollet says that LLMs are like databases

[00:38:34] Next frontier of language models

[00:42:04] Different modes of understanding in LLMs

[00:47:04] LLMs are the new extended mind

[00:50:03] Is language the next interface, and why might that be bad?

References:

[Balestriero] Spine theory of NNs

https://proceedings.mlr.press/v80/balestriero18b/balestriero18b.pdf

[Delétang et al] Neural Networks and the Chomsky Hierarchy

https://arxiv.org/abs/2207.02098

[Fodor, Pylyshyn] Connectionism and Cognitive Architecture: A Critical Analysis

https://ruccs.rutgers.edu/images/personal-zenon-pylyshyn/docs/jaf.pdf

[Chalmers, Clark] The extended mind

https://icds.uoregon.edu/wp-content/uploads/2014/06/Clark-and-Chalmers-The-Extended-Mind.pdf

[Melanie Mitchell et al] The Debate Over Understanding in AI's Large Language Models

https://arxiv.org/abs/2210.13966

[Jay Alammar]

Illustrated stable diffusion

https://jalammar.github.io/illustrated-stable-diffusion/

Illustrated transformer

https://jalammar.github.io/illustrated-transformer/

https://www.youtube.com/channel/UCmOwsoHty5PrmE-3QhUBfPQ

[Sandra Kublik] (works at Cohere!)

https://www.youtube.com/channel/UCjG6QzmabZrBEeGh3vi-wDQ

Kategorier
Förekommer på
00:00 -00:00