I have been writing code for over 15 years and have worked with relatively complex distributed systems. These are my own experiences comparing the web3 world to traditional software engineering ๐งต
I built a thing for the kids; you are all welcome to use it
Voice recommendations for movies/shows/games/books/recipes.
Feedback is always welcome
https://t.co/SEn8YpXclU
AI is amazing, but I think a lot of people underestimate how hard it is to get real value from it.
The models are usually not the difficult bit.
The hard bit is the messy data, vague intent, old systems, weird edge cases, and making something that actually fits into how people..1
Finally, if you know someone whoโd find this useful, or someone working on practical AI problems where this kind of thinking would help, please share it with them.
https://t.co/SEn8YpXclU
AI is amazing, but I think a lot of people underestimate how hard it is to get real value from it.
The models are usually not the difficult bit.
The hard bit is the messy data, vague intent, old systems, weird edge cases, and making something that actually fits into how people..1
It started as something for my family, but itโs also the kind of ai work i enjoy most: taking messy real-world intent and turning it into something useful.
Iโd really love feedback on it...5
@sh_reya Agreed. I'm finding NL to SQL significantly better for both accuracy and relevance
Ofcourse it's significantly harder than tradtionaly RAG especially in large orgs with TBs of data with datasets and relationships aren't well defined
what makes LLM frameworks feel unusable is that there's still so much burden for the user to figure out the bespoke amalgamation of LLM calls to ensure end-to-end accuracy. in https://t.co/0rYHaAKXHv, we've found that relying on an agent to do this requires lots of scaffolding
@sh_reya is one of the pioneers in this space. Another brilliant thread
If you are doing any work with LLMs and unstructured data then be sure to give her a follow!
what makes LLM frameworks feel unusable is that there's still so much burden for the user to figure out the bespoke amalgamation of LLM calls to ensure end-to-end accuracy. in https://t.co/0rYHaAKXHv, we've found that relying on an agent to do this requires lots of scaffolding
Imbue has officially joined @GoogleStartups Web3 startup program marking a significant milestone in our innovation journey.
Backed by a generous $200k in credits we are ready to #MakeWeb3WorkForYou
https://t.co/LQdSLJ7UNJ