Quoted By:
Would Neuro and Vedal survive an AI winter?
>AI winter is a period where AI research reaches a wall; the current methods and paradigm prove to not be good enough and it's apparent that we're barking up the wrong tree. Thus investment in AI dries up.
Here's the scenario I'm working under: In Feb 2025 it turns out that the AI fad is over, LLMs will never be able to count how many r's are in strawberries, never will be able to reason etc. Perhaps ML in general is insufficient. Investment dries up. Everybody moves onto quantum or gene editing, aside from a persistent few who will still slowly progress the field. It will only be until Mar 2028 that these persistent few manage to revitalise the field by introducing some new paradigm. (Azure Voice API and the other APIs that Vedal uses will continue to exist.)
I think this would massively shrink the Swarm to 3view.
Vedal's progress would stagnate massively due to not being propelled by the entire AI community. He would fail to finish among us, because multimodal speech, understanding the rhythm of conversations and reasoning would remain unsolved problems. He will fail to give Neuro spatial reasoning/awareness, because that is an unsolved problem (thus blocking off the potential for VR exploration). Neuro's agentic abilities won't advance. Neuro's won't be able to control robots in a complex, granular way (as opposed to simple joystick control). Neuro's vision wouldn't advance.
Thus I reckon he will not be able to provide any novel content for 3 years. The devstreams will be rather sparse and hollow, and as the viewcount for those drop, the viewcounts for other streams drop in turn.
Lucrative sponsorships dry up. The offers to buy Neuro disappear overnight. Anny will leave, because Vedal couldn't pay the rent. pb and tomyomy move on. His harem disappears, as the (former) Neuroverse continue to chug along unaffected. Mr. Minibot withdraws his recommendation that Ellie should marry Vedal.
And then, after a gruelling 3 year slog, he sees the end of the winter. He joyously announces this to the 350 people left.
But after a while, whilst lying in bed, he realises that the new paradigm is incompatible with LLMs and thus Neuro cannot make the transition.
There might be some opportunities though.
Vedal could buy up the now significantly cheaper AI datacentre gpus, after the datacentres shut down. He could use these GPUs to make Neuro very fast, perhaps even fast enough to play real-time games with the game API. He could also use it for training/R&D.
Maybe he could get some AI researchers to help upgrade Neuro to SOTA. (e.g. Get a former Anthropic engineer to give 'Computer use' to Neuro so she could use GUIs, finally develop a good 11labs-tier voice)
