nafyi.

What Comes After Transformers?

The Data Exchange ·

Zuzanna Stamirowska calls RAG, context, and memory bottlenecks in 2023 and 2024 for what she describes as 'AI with no ceiling, if you wish.' She recalls realizing that reasoning should not just be linguistic, and says latent thinking can help solve constraint-satisfaction problems and math with 'tiny, tiny models.'. Lisez 3 points de vue avec leurs éléments à l’appui et les liens vers les sources.

Ben LoricaZuzanna Stamirowska
Lire les points clésOuvrir Éditeur (s’ouvre dans un nouvel onglet)

En un coup d’œil

  • RAG, context, and memory were bottlenecks in 2023–2024

    Stamirowska says the team saw RAG, context, and memory as bottlenecks in 2023 and 2024. They were not efficient or sufficient for what she describes as 'AI with no ceiling, if you wish.'

    Lire le moment probant · Publisher transcript paragraph 7

  • Stamirowska says reasoning shouldn't just be linguistic

    Stamirowska says, 'We realized that reasoning shouldn't just be linguistic.' She says verbalizing things in chain-of-thought is not optimal: it involves 'gluing it on top of a transformer,' and transformers are naturally ill-suited to dealing with time and learning from experience.

    Lire le moment probant · Publisher transcript paragraph 8

  • Latent thinking can help solve constraint-satisfaction and math problems

    Stamirowska says latent thinking can help solve constraint-satisfaction problems and math with 'tiny, tiny models.'

    Lire le moment probant · Publisher transcript paragraph 37

Moments clés3

Extraits courts et attribués, avec le contexte permettant de les vérifier. La conversation complète reste chez son éditeur.

Publisher transcript paragraph 7 (s’ouvre dans un nouvel onglet)
post-transformer architecture motivation

RAG, context, and memory were bottlenecks in 2023–2024

Extrait original

But very concretely, what prompted us to start Pathway and to start working on post-transformer architectures was that we were looking forward to reasoning becoming a thing. We have a very strong background in dealing with the dimension of time in machine learning—any sort of online learning. And we saw RAG, context, and memory, and we’re talking about 2023 and 2024, as bottlenecks. They were not efficient or sufficient to unlock AI as we would like—AI with no ceiling, if you wish.
Zuzanna Stamirowska·
Publisher transcript paragraph 8 (s’ouvre dans un nouvel onglet)
reasoning modality

Stamirowska says reasoning shouldn't just be linguistic

Extrait original

Specifically, we’re talking about continual learning, memory, and reasoning first. And as we moved through it, we realized that reasoning shouldn’t just be linguistic. Verbalizing things in chain of thought is not the optimal way to do reasoning. With chain of thought, we’re essentially gluing it on top of a transformer, and transformers are naturally ill-suited to dealing with time and learning from experience.
Zuzanna Stamirowska·
Publisher transcript paragraph 37 (s’ouvre dans un nouvel onglet)
latent reasoning benefits

Latent thinking can help solve constraint-satisfaction and math problems

Extrait original

Latent thinking has huge benefits. It can help us solve constraint-satisfaction problems natively, like what we’re showing. We’re solving constraint-satisfaction problems and math with tiny, tiny models. And then there’s cost efficiency. There are just insane cost gains coming from latent reasoning.
Zuzanna Stamirowska·

Source et méthodologie

Ces points de vue renvoient à leurs sources originales. Les reformulations sont signalées et ne sont pas des citations mot à mot.

Ouvrir la transcription ou les documents sources (s’ouvre dans un nouvel onglet)Signaler un problème