Dan Romero on Warpcast

Dan Romero pfp

LLMs are neat. But seem like a sustaining technology vs. a disruptive technology.

14 replies

0 recast

0 reaction

Dan Romero pfp

More market cap for Big Tech vs. startup decacorns. Notwithstanding the huge fundraisers for the big tech proxy mercenaries (OpenAi, Anthropic, Hugging Face).

2 replies

0 recast

0 reaction

Gambid pfp

counterpoint: if you just keep scaling up data and compute they will at some point be quite disruptive as they pick up more emergent abilities

1 reply

0 recast

0 reaction

Chu Ka-Cheong pfp

Agree. LLM is surely revolutionary, but it isn’t necessarily a disruptive technology like internet that can create new market. LLM is so depends on access to large amount of high quality data, large GPU farm, and early access to latest model. These criteria favor incumbents (like Microsoft) more than startups.

1 reply

0 recast

0 reaction

Jason Crawford pfp

@jasoncrawford.eth

I think it depends on what *product* is built with it and how. You could definitely build disruptive products with LLMs. An LLM-based therapist, e.g., might not be as good as a person but would be much cheaper. Classic profile of a disruptive product

0 reply

0 recast

0 reaction

Ben Scharfstein pfp

Ben Scharfstein

I don't think it's a "vs" kind of discussion. For some industries they'll be hugely disruptive and for others they'll be sustaining tech.

0 reply

0 recast

0 reaction

Michael Pfister pfp

Michael Pfister

I was talking to @jj today about this and he convinced me that LLMs will be like pre/post internet level disruption. People just haven’t spent enough time learning how to use LLMs effectively. Worst case, everyone on earth just got a $20/mo polyglot personal assistant with 20 different Bachelor degrees

1 reply

0 recast

0 reaction

Avi 💙 pfp

Also @mog sent me this today. And it’s nuts! https://twitter.com/ruima/status/1636042033956786177

0 reply

0 recast

0 reaction

Avi 💙 pfp

Depends on the context but I say disruptive - Journalism: rapid content generation w/ help of human editors - streamline contract reviews & legal research, disrupting junior lawyer roles - Healthcare: analyze medical research & assist in diagnostics, enabling meds to focus on patient care - Language translation

0 reply

0 recast

0 reaction

Nicholas Charriere pfp

Nicholas Charriere

I don’t think so but I have little to back it up. Intuitively feels like one of the most important breakthroughs because of the distribution of the interface and the low barrier to entry to learn it.

0 reply

0 recast

0 reaction

Joe Petrich 🟪 pfp

Joe Petrich 🟪

Agree, but I think the sustaining technology can essentially be a multiplier, so that the timeline on which disruptive tech overtakes dominant tech is accelerated. And, if smaller orgs are able to adopt the sustaining tech faster, then it would increase the relative rate of improvement of the disruptive tech too.

0 reply

0 recast

0 reaction

osama pfp

llms are an autocomplete feature at best for products with existing distribution on a 3-5year horizon. (poor man’s weekly held opinion)

2 replies

0 recast

0 reaction

Syed Shah🏴‍☠️🌊 pfp

Syed Shah🏴‍☠️🌊

Disagree. Death by a thousand cuts for large corporations that are inherently very inefficient. Could be a massive restructuring across the board, but def in tech.

0 reply

0 recast

0 reaction

Renjith Nair pfp

At the end of it this going to be a training dataset game . It’s very hard to have sustained algorithmic advantage . Eventually LLM are going to be utility and even IBM will have one .

0 reply

0 recast

0 reaction

Theodor Marcu pfp

agreed! feels like it's going to be 80/20 incumbents vs startups but mobile was the same! so still plenty of opportunity if you know where to look. my hot take is that I actually think startups in consumer AI are probably going to do better than the myriad of b2b SaaS / dev tools startups built around OpenAI APIs

0 reply

0 recast

0 reaction