Localized reading
Cada vez mais acho que o software do futuro vai combinar código para o fluxo de controle com pequenos programas neurais para julgamentos “imprecisos”. É isso que venho explorando com o ProgramAsWeights, que responde de onde vêm esses programas neurais: eles são “compilados” a partir de descrições em inglês. Por exemplo, combinei 30 programas neurais com uma árvore de decisão para criar um assistente para um site de cursos que parece um chatbot, mas roda localmente. Cada pequeno programa lida com perguntas como “para qual especialista isso deve ser encaminhado?”, enquanto o código controla todo o fluxo…
Original post / EN
I increasingly think future software will combine code for control flow with small neural programs for "fuzzy" judgments. That's what I've been exploring with ProgramAsWeights, which answers the question of where those neural programs come from: they are "compiled" from English descriptions. For example, I combined 30 neural programs with a decision tree to build a course website helper that looks like a chatbot but runs locally. Each small program handles a question like "Which specialist answerer should this be routed to", and code controls the overall flow. You can try building with it here: https://t.co/Bwgcn362bc
- views
- 13,5 mil
- likes
- 171
- saves
- 140
- reposts
- 21
Links and projects
Quoted post
Diogo Almeida @CompleteSkeptic
After co-inventing ChatGPT, I kept asking myself: why have superhuman chat models not led to AGI? I’ve spent the last 2 years in stealth building a new way to train models (RLCD), and a new type of frontier AI model that we are releasing today: Jev • 20-200x faster • 40-400x https://t.co/JSybNG2BKJ
Ver original ↗Fonte
This independent archive preserves a public post with attribution. Text, media, account details and trademarks belong to their respective owners.
