MCLIDE @MCLIDE

#AI at IA this Wed's #DWeb Meetup: come hear how decentralization & open data may be key to #AI's future w/ talks by @StabilityAI, @OpenAI, @AiMozilla & Gun Database founder @marknadal. Food, drinks, networking & deep learning about #AI: eventbrite.com/e/distributed-…

1 11 38 17K 2

0 0 1 331 0

View Details

Paul Graham @paulg

3 years ago

But the nerdy enthusiasts will build stuff for them on top of AI as well. Which will also presumably have higher adoption rates, since so many end users will already be believers.

15 16 458 63K 15

View Details

MCLIDE @MCLIDE

3 years ago

@Suhail Alternatively, first movers leave everybody else in the dust.

0 0 1 94 0

View Details

MCLIDE @MCLIDE

3 years ago

@cesmithjr @OfficialLoganK @Suhail "long tail" - I see what you did there.

0 0 0 32 0

View Details

MCLIDE @MCLIDE

3 years ago

@OfficialLoganK @Suhail Second movers get to eat the cheese while the first movers lie with their back broken... appreciate those willing to be first movers.

0 0 0 82 0

View Details

MCLIDE @MCLIDE

3 years ago

@corcoranwill @igorsyl @TO @johnennis @paulg @TerjeNorderhaug No problem, @TerjeNorderhaug is my private account.

0 0 1 34 0

View Details

MCLIDE @MCLIDE

3 years ago

@corcoranwill @igorsyl @TO @johnennis @paulg @TerjeNorderhaug Hah! I leveraged Common Lisp for my first startup in grad school, having independently had the same insight as @pg in "Beating the averages". In 1995 we shipped the first social web platform. terje.norderhaug.name/interaction

0 0 3 104 0

View Details

MCLIDE @MCLIDE

3 years ago

@dotta @OfficialLoganK Functional languages should benefit generative AI by keeping computation/semantics local. Unfortunately, due to the selection of training examples, current models are biased in favor of imperative languages.

0 0 0 22 0

View Details

MCLIDE @MCLIDE

3 years ago

@dotta @OfficialLoganK If you use lisp you can almost fit the whole program into the context window - if it weren't for that a dash `-` count as a token, so symbols with dashes will quickly eat up the limited number of tokens available.

0 0 0 30 0

View Details

MCLIDE @MCLIDE

3 years ago

@johnennis That's because you're an entrepreneur. There's an abundant supply of improvements that can be made, particularly in the digital domain.

0 0 0 32 0

View Details

MCLIDE @MCLIDE

3 years ago

@OfficialLoganK The limitations on LLM context favor languages that use the least tokens, whether due to dense code or effective encoding.

0 0 0 43 0

View Details

MCLIDE @MCLIDE

3 years ago

@OfficialLoganK Pre-trained language models seem to have hidden representations of abstract syntax trees. I am leveraging this in a language with a syntax close to the AST.

0 0 0 57 0

View Details

MCLIDE @MCLIDE

3 years ago

@TO @paulg You may enjoy the brilliant multi-talented @lisperati

0 0 1 45 0

View Details

MCLIDE @MCLIDE

3 years ago

@fluidvoice @TO @paulg M for Mac. It was a remote REPL developer tool I made back in the aughts: cliki.net/MCLIDE

0 0 1 81 0

View Details

MCLIDE @MCLIDE

3 years ago

@TO Lisp is also great for programs that process Lisp code. That's why the Clariform linter & formatter for @clarity_lang smart contracts is written in Lisp: github.com/PromptECO/clar…

0 0 0 10 0

View Details

MCLIDE @MCLIDE

3 years ago

@TO People new to programming may be quicker to pick up Lisp syntax than many developers already set in their ways from programming in other languages. Lisp has a minimalistic syntax that is easy to introduce.