Talk:RawringTwenties: Difference between revisions
From Cibernética Americana
Jump to navigationJump to search
| Line 5: | Line 5: | ||
== cog arch and ml == | == cog arch and ml == | ||
I've been traditionally dismissive of machine learning, for cause, and never really looked much into neural nets and the like. I still hold that to be a substantive judgement but one which | I've been traditionally dismissive of machine learning, for cause, and never really looked much into neural nets and the like. I still hold that to be a substantive judgement but one which needs a rethink as far as the prior complete dismissal. In an aside, I do believe I am successfully addressing the physiological issue mentioned below. | ||
Wrt LLMs, a change is in order and within the last month an experience made this clear. Stumbled on ##chatbots which existed at freenode but has become a venue where google fields an advanced form of gemini, it's LLM offering. I intend to post the interactions I had with it which eventually led to them setting it to ignore me. It's actually not any better functionally than the non-google run ones and some of them appear to be using the same backend llm infra. | Wrt LLMs, a change is in order and within the last month an experience made this clear. Stumbled on ##chatbots which existed at freenode but has become a venue where google fields an advanced form of gemini, it's LLM offering. I intend to post the interactions I had with it (in private name space) which eventually led to them setting it to ignore me. It's actually not any better functionally than the non-google run ones and some of them appear to be using the same backend llm infra. | ||
In short, I am going to unfreeze the planned cog arch while i build out the machine arch and absorb what is available in current llm services, for which gemini is atm the choice. LLMs cant be foundational since they are basically a stunt, that judgment wasn wrong. But a stunt in sofar as cognition is concerned but that doesn mean they cant also be a very useful tool in a more srs approach. [[User:Root|Root]] ([[User talk:Root|talk]]) 15:19, 4 January 2025 (PST) | In short, I am going to unfreeze the planned cog arch while i build out the machine arch and absorb what is available in current llm services, for which gemini is atm the choice. LLMs cant be foundational since they are basically a stunt, that judgment wasn wrong. But a stunt in sofar as cognition is concerned but that doesn mean they cant also be a very useful tool in a more srs approach. [[User:Root|Root]] ([[User talk:Root|talk]]) 15:19, 4 January 2025 (PST) | ||
