Hacker News Reader

Comment by yeldarb | Hacker News Reader

> But most of the engineers I personally know are sort of squinting at it and thinking, “Is this another crypto?” Even the devs at Sourcegraph are skeptical. I mean, what engineer isn’t. Being skeptical is a survival skill.

> Remember I told you how my Amazon shares would have been worth $130 million USD today if I hadn’t been such a skeptic about how big Amazon was going to get, and unloaded them all back in 2004-ish.

If you buy into the hype & think this wave is going to be equivalently (or 10x) as impactful as cloud computing, what's the equivalent to buying Amazon stock in 2003?

View on HN

Buying shares in exactly the right company that will conquer the market, while avoiding the 99% of companies that look just as good right now but will fall by the wayside somewhere in the coming decades.

> If you buy into the hype & think this wave is going to be equivalently (or 10x) as impactful as cloud computing, what's the equivalent to buying Amazon stock in 2003?

About ten years ago, a big VC partner gave a talk at Stanford. In his talk he asked the audience which company would be “the next Google”. After hearing a few suggestions, he said the company hadn’t even been started yet hence nobody could really know.

Changes are happening so fast now, that I wouldn’t be surprised if 6 months to 1 year from now, something comes out that kills GPT.

It’s very hard right now to tell who the winners will be.

For now, if I had to choose (this is not financial advice), maybe betting on MS and NVIDIA could pay off in the short-midterm. But 2-5 years from now? Impossible to know.

I just want to second the (not financial advise) line, the market dynamics now are not favorable to taking these sorts of bets, particularly since both companies got run up in the pandemic.

> something comes out that kills GPT.

Can't even assume the LLM providers will be the big winners. LLMs can be replicated: https://pub.towardsai.net/meet-alpaca-stanford-universitys-i...

for much cheaper than they can be built.

If the laws prohibit it (as they do in the case of Alpaca) then it doesn’t really matter how cheaply they can be built: the resulting AI won’t be able to be significantly monetized.

But a lot of companies could just use it anyway.

Use it now, pay for the consequences later. Not condoning it, just pointing it out.

Things are moving so fast, that in 1-2 months there will be an open source or free version of a GPT-3 level LLM out. At that point they can swap out their illegal LLMs for the free/open one and done.

OpenAI knows this and that’s why they are working so hard in multi-modality - to be able to keep their edge.

I’m not going to argue, and I think I may have made a similar point if I had come across my comment, but that’s why I chose to say “significantly”. Any startup could certainly “use it now”, but the higher your valuation the higher the scrutiny and history is littered with companies that had to close shop once it was discovered they were cheating. You can probably get away with it when small, probably weather the storm if you’re big, but in between is a danger zone.

You are right.

Being illegal didn’t stop Uber, nor Airbnb, nor a bunch of other companies and services.

If you grow fast enough, the growth becomes so valuable that no one really wants to (or maybe can?) stop you.

If you are small, nobody cares.

And like you said, in the middle is a lot more dangerous.

If the author had bought Bitcoin back in 2010 instead, and held them through all the chaos to this very moment, they would have approximately 20000x their initial investment. I'm not sure what, if anything, that says about crypto being a fad, rather than just cherry-picking historical examples.

I think what he meant is that it didn’t change software engineering. Not that many people write code that touches cryptocurrency in some way.

If you worked on cryptocurrency software, how useful is that experience now?

It's not a out making money, it's about making skills. You need tensor$ to make money. If you aren't using it and learning it daily you're going to be left behind in the knowledge wealth department.

Get in late and you'll be catching up with experts on when and when not to use it, how best to use it, and how best to weed out the bullshit that it intersperses throughout the gold.

There are more ways to skin this cat than Copilot alone. For example, asking ChatGPT questions, or plugins like CodeGPT.nvim, or some other future tool that you wouldn't otherwise be considering.

Remember how strong Google-fu was almost a superpower (still somewhat is, to be honest)? We're currently in the PageRank days of ML.

You know who made even more money than people who knew how to use Google? The people who owned Google. That is what GP was asking about, how to become someone who owns the LLM that everyone will ask their questions to.

I address that in the first paragraph. You need a lot of money (tensor ops) to make money using LLMs. You're going to depend on the big players no matter what.

I don't think LLMs are good enough yet for embedding into apps - or we (tech employees) don't have enough experience with them to be able to innovate them into an end-user product, but I'm open to being wrong about that.

I agree that you need to dive in if you want to take advantage of current opportunities.

On the other hand, some skills could go obsolete pretty quickly. People new to the game will be able to avoid previous mistakes and skip the old stuff. It seems unlikely that future college grads will be missing out due to lack of experience?

Unfortunately, it seems like its the big boys. Open AI is having its moment, but otherwise seems to be about the megas like Google & Microsoft.

Amazon destroyed the used book business in cities across the Western world; I have never purchased a book from Amazon dot com. Is the world a better place with your $130 million fantasy money? at what cost

Reply on news.ycombinator.com