context
8
ctx:discord/blah/papers/8Source document
full textpapers-8
text/plain3 KB
doc:agent/papers-8/6fb41b21-3cff-4525-91b8-47c46c7190e4[2026-03-27 03:25] traves_theberge: are you talking about neural oscillations? [2026-03-27 03:25] lisamegawatts: yes, the theta gamma etc [2026-03-27 03:26] lisamegawatts: og oscillator [2026-03-27 03:27] lisamegawatts: (files: message.txt) [2026-03-27 03:29] lisamegawatts: (files: message.txt) [2026-03-27 03:31] lisamegawatts: the antiforgetting is not attention for my purposes it ismemory i am tired of telling claude the same think, and it only has attention now [2026-03-27 03:31] traves_theberge: # Geometric Resonance in Multimodal Learning: Topology-Conditioned Coordination in Oscillator Networks i love everything about this sentence. [2026-03-27 03:32] lisamegawatts: so in this case i am trying to baby step always remember, and recall no slip off the context, no limit [2026-03-27 03:32] lisamegawatts: continual learning [2026-03-27 03:32] traves_theberge: Feynman-splan this to me lol [2026-03-27 03:33] lisamegawatts: we proved basically we can lear new ones and not forget old, if the field can do that too which i think it can that would be the way [2026-03-27 03:34] lisamegawatts: the multi model we gt rid of even concept of tokens and everything is byte > qkps> wave, and in that format any and all bytes can be added [2026-03-27 03:35] lisamegawatts: so we made one that could do i think wave forms, text, video, image, and wav files [2026-03-27 03:35] lisamegawatts: just all bits, it doesn't need to specify, and they learn where to go on their own [2026-03-27 03:36] lisamegawatts: the idea was also what if any input all signal no label [2026-03-27 03:37] traves_theberge: how? Where are the receivers? [2026-03-27 03:37] lisamegawatts: er i think i have over 300 papers one per experiment [2026-03-27 03:38] lisamegawatts: oh ha well one of the oscillator models i rebuilt software defined radio lol [2026-03-27 03:38] lisamegawatts: like all of it, when i made that telegraphers comment [2026-03-27 03:39] lisamegawatts: i added telegrapher equations and sigint, parity fec, multitapers, bandpass... [2026-03-27 03:39] lisamegawatts: like og radio [2026-03-27 03:40] lisamegawatts: but er for the wave, the geometry wants a wave and bits are the og wave to discrete, it is reversible [2026-03-27 03:40] lisamegawatts: like a bit is made to be the translation from waves in wires, i flip it and my bits become waves [2026-03-27 03:42] rolandnsharp7643: <@1211062099137265723> Have you tried causal generation? any speech continuation for example from pure oscillators? [2026-03-27 03:42] lisamegawatts: yes [2026-03-27 03:42] rolandnsharp7643: awesome [2026-03-27 03:42] lisamegawatts: the probe i sent you, we used it to see the structuresa it learned, like it would respond different to spaces before learning english [2026-03-27 03:42] lisamegawatts: richard got it to talk but not great [2026-03-27 03:43] rolandnsharp7643: it's talking ! mine said woo woo woo [2026-03-27 03:43] lisamegawatts: 🙂
Facts in this context
Grouped by subject. Each subject links to its full article.
Lisamegawatts18 factsex:lisamegawatts
| affirmedTopic | Neural Oscillations |
| claimsProofOf | Catastrophic Forgetting Solution |
| definesPurpose | Antiforgetting |
| describesGoal | Baby Step Remembering |
| distinguishesConcept | Antiforgetting |
| expressesFrustrationWith | Claude |
| mentionedHistoricalComment | Telegraphers Comment |
| performedOperation | Bit to Wave Conversion |
| possessionCountEstimate | 300 |
| proposedIdea | Unsupervised Signal Processing |
| rdfs:label | lisamegawatts |
| rdf:type | User |