Lampinen calls out neurosymbolic camp for walking back internal-symbolism claims
AndrewLampinen · x · 2026-10-03
DeepMind researcher Andrew Lampinen debates a neurosymbolic proponent, arguing that symbolic processes generating training data (like math labels) don't require a neurosymbolic approach—RL can achieve the same. He adds that claims models must have internal symbolic components were only walked back once they became untenable, in a pointed spat over whether LLMs need real symbolism.
More from AGI Musings
- Burkov: Recursive AI Self-Improvement Will Look Funny at 3 Months per Training Run — burkov · 2026-10-03
- Cloudflare dev Sunil Pai says machines can be conscious — and it's not a dramatic belief — threepointone · 2026-10-03
- Why AI can't solve the mystery of time: training presupposes the very clock it must explain — johnseach · 2026-10-03
- arXiv trends suggest AI uplift hits quantum physics within a year, all physics by late 2029 — cephaloform · 2026-10-03
- Fifty-Year AI Dream Finally Built — Yet Its Creators Insist They Made God — bradneuberg · 2026-10-03
- X is mostly capital owners, not artists: why the AI discourse has no other side — moonsandhues · 2026-10-03