aik

When did hacker news become laggard-adopter/consumer-news.

Cal is a consumer of AI - interesting article for this community, but not this community. I thought hacker news was for builders and innovators - people who see the potential of a technology for solving problems big and small and go and tinker and build and explore with it, and sometimes eventually change the world (hopefully for the better). Instead of sitting on the sidelines grumbling about that some particular tech that hasn’t yet changed the world / met some particular hype (yet).

Incredibly naive to think AI isn’t making real difference already (even without/before replacing labor en masse.)

Actually try to explore the impact a bit. It’s not AGI, but doesn’t have to be to transform. It’s everywhere and will do nothing but accelerate. Even better, be part of proving Cal wrong for 2026.

We’re a regular old SaaS company that has figured out how to add massive value using AI. I am making no statements about valuations and bubbles. I’m actually guessing there is some bubble / overhype. That doesn’t mean it isn’t still incredibly valuable.

I have no idea if OpenAI’s valuation is reasonable. All I’m saying is I’m convinced the demand is there, even without AGI around the corner. You do not need AGI to transform countless industries.

And we are profitable on our AI efforts while adding massive value to our clients.

I know less about OpenAI’s economics, I know there are questions on whether their model is sustainable/for how long. I am guessing they are thinking about it and have a plan?

Hard disagree. You don’t need AGI to transform countless workflows within companies, current LLMs can do it. A lot of the current investments are to help with the demand with current generation LLMs (and use cases we know will keep opening up with incremental improvements). Are you aware of how intensely all the main companies that host leading models (azure, aws, etc) are throttling usage due to not enough data center capacity? (Eg. At my company we have 100x more demand than we can get capacity for, and we’re barely getting started. We have a roadmap with 1000x+ the current demand and we’re a relatively small company.)

AGI would be more impactful of course, and some use cases aren’t possible until we have it, but that doesn’t diminish the value of current AI.

Agreed.

Confused here, is your argument here that OpenAI is not responsible for any innovation when it comes to LLM tech today? I’m curious about why you so strongly want to believe that?

Nobody knew that scaling transformer architecture would lead to the emergent intelligence we see today. Among other things, OpenAI did R&D for years on that. Also the only situation where this could true is if Google knew that LLMs could lead to this intelligence and decided to not make it happen, (along with every other tech company now that is furiously trying to catch up to OpenAI), which is absurd.

The money and compute is not the innovation. The LLM models and associated tools are, which is work by OpenAI employees and teams, not Microsoft employees/teams.

I think this perspective is probably only true for 0.001% of people that actually follow Sam closely and are not optimistic about AGI and like to throw their opinions around. The superficial stuff. The rest don’t care to even know who Sam is and don’t care to assume motive.

It’s very likely they’ll bounce back. I’d rather OpenAI continue to innovate and push the industry forward as they have been. Haven’t seen much of that from Microsoft, so heavily disagree with you there. Prefer to focus on the actual product of the company not the personalities of the people there or armchair assumptions on the vibes of the culture.

Why? It’s not firewalled.

1. “The heat death of the universe” is my favorite HN comment of the decade.

2. The heat death of the universe does not mean one gigantic black hole. I’m just a hobbyist but my understanding of the theory is that black holes will continue to form, but through Hawking radiation, they eventually radiate out all their energy until it is all dispersed, ultimately leading to uniformity across the entire universe, max entropy, where “work” can no longer take place.

(It is an interesting question then whether information is actually destroyed through Hawking radiation?)

Curious what model the dentist bot is running on? Tried it out, was surprisingly good, though eventually it contradicted itself (booked a slot it said previously was not available). (I get that’s the programming but am curious especially given the latency is really great).

Completely disagree. You’re not making a phone call in most cases for entertainment purposes. If the options are wait in line for 20 minutes or speak to an actually useful bot, I would take the latter in 100% of cases.

I think standard units are more common even if proportions. Not sure who speaks in bananas.

Have to consider cost for all of this. Big value of RAG already even given the size of GPT-4’a largest context size is it decreases cost very significantly.

Not entirely sure I believe everything here. Curious how it plays out. Lot of narrative-affirming content.

“In the Examiner article, O’Reilly made reference to a “scam bot [that] had a blue check mark, meaning that, unlike me, it pays money every month to Elon Musk’s vastly indebted and unprofitable platform, a situation which would greatly disincentive his company taking proactive measures to weed them out”.

Is the argument here that someone is creating an army of paid accounts and spamming with them? And this is a major source of bots? Is this actually happening? I think that is probably unlikely given how expensive it would be?

Not sure if you’re technical but the only thing I have to say to this is: Tinker with it yourself. Try different experiments. I’ve built a ton of tools at this point with AI, some have not been very useful in the end and others have made me significantly more productive and effective.

In terms of error rate: gpt 3.5 had a high hallucination rate that made use cases fairly narrow. It then got faster which opened up some more use cases. Then gpt 4 came out that had a significantly smaller hallucination rate which opened up a gigantic number of additional possibilities. And had a larger context window and output size that made it significantly more useful. Then it got faster with an even larger context size… each of these iterative improvements just continue to add more and more possibility in a gigantic range of cases that have literally never existed before.

1. We know they have more big breakthroughs already that have not been released. 2. We know the current tech can keep scaling. They have not hit a limit with the current approach yet.

Given gpt-4 is already ridiculously useful and we’ve barely scratched the surface, it makes complete sense to me. More capacity + faster gpt responses unlocks massive amounts of more potential/use cases.

I am guessing an incredibly talented team that is incredibly networked and incredibly well funded and proven agile in the tech hub of the world can find hardware experts. Don’t know why anyone would bet against that.

Said the horse factory when automobiles were being built.

I agree some of these are fluff. However many here are great to have basic knowledge of for specific scenarios one may run into and then know where to dig deeper when needed.

Agree though to marry being a ferocious reader / learner with being a do’er is the way to go. Personally I did not grow up with an opportunity to develop an intuition about business and self-motivated action. Has taken a lot of work to move that direction over the years.

It’s very likely this person doesn’t care about the credentials but rather just wants to gain the knowledge to do something themselves.

I more or less did this for years by reading a ton of books + many years of HN / articles / interviews / videos + startup weekend events + tried starting things + coding hobby for years + enterprise consulting work, and now I’m fairly well equipped to co-lead the company that I do.

The first 5+ years of doing this I debated attempting to go to a prestigious school for an MBA, primary reason (95%) was for the networking opportunity (meeting great likeminded or driven people), (0% because I would’ve been able to tell people I had an MBA from somewhere for job purposes, the snobbery from that would have been a detriment often instead I believe), and 5% for the education which I think has value but when it’s free and I had the motivation there was no reason to pay $100ks for that. At this point this company is doing fairly well so although I would love to meet more driven and likeminded people, I’m not sure it the best path to do so anymore for me.

Said people before nearly every discovery in physics led to new technology.

2 questions:

1. How does this perspective affect you today? Is it debilitating? Depressing? Affect your wellbeing and productivity?

2. What do you believe is the probability of what you say here? Estimate with a percentage.

Does the unknown associated to the percentage in #2 make the perspective rational and helpful?

And encourage innovation it does.

Most initiatives are not purely good without any downsides. A principle to consider is: Just because there are some negative aspects sometimes, do those outweigh the good? In this case, how many negatives actually come from patent law? (Btw, a monopoly most often is not the result - there are often many solutions to a similar problem/need.). And consider the side of the creator (which I’ve been finding is surprisingly rare in hn) - creating a company and product is already incredibly hard, how much harder do we want to make that? What and who are we sacrificing?

Maybe we just adjust the patent law some. Decrease the number of years? I’m sure it could be better.

Not public but internally I wrote a tool to help us respond to RFPs. You pass in a question from a new RFP and it outputs surprisingly great answers most of the time. Is writing 75%+ of our RFP responses now (naturally we review and adjust sometimes and as needed). And best of all it was very quickly hacked together and it’s actually useful. Copied questions/answers from all previous ones into a doc, and am using OpenAI embeddings api + FAISS vector db + GPT-4 to load the chunks + store the embeddings + process the resulting chunks.

More often he has expressed how well it works and how great it is.

Agreed. Thank you Tesla for helping bring about a better standard here!

Raw text from a website including header text and footers and links and images etc is very dirty stuff.

Do you have an example of where these methods still produce good summaries? Eg if you adjust how re-computation of self-attention in autoregressive decoding / between token generations works to significantly decrease the amount of computation needed?

Am not justifying what OpenAI did, but nobody is stopping ByteDance from doing what OpenAI did. They can also use the world’s information. Instead, since OpenAI has “cleaned” the data, they are trying to use OpenAI’s cleaned dataset. After OpenAI spending endless amounts of money on that, am not surprised they don’t want others to steal their “cleaned” dataset.