Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

GPT-2: Too Dangerous To Release (2019)

naokishibuya.github.io

Post reply on HN
Print view View original

111–120 of 139 posts

Re: GPT-2: Too Dangerous To Release (2019)

#111

Post by senordevnyc »

malfist wrote: post #87
Qhemlomo wrote: post #21

Earlier quoted context omitted.

How is this a gimmick? It changes my whole profession on a level i couldn't even imagine how we would 'solve' software engineering.

> It changes my whole profession on a level i couldn't even imagine I assure, it doesn't.

It already has!

It’s mind-boggling that anyone could deny this in mid-2026. Virtually every software engineer I know is no longer writing the majority of their code. Many are not writing any code, myself included. And I’m a staff engineer with 20 yoe, formerly at big tech, and now building a (profitable) SaaS of my own. The way I work is wildly different from a year ago.

And no one is going back.

Re: GPT-2: Too Dangerous To Release (2019)

#112

Post by autoexec »

panzi wrote: post #25
Qhemlomo wrote: post #19

Earlier quoted context omitted.

I don't want to stop progress just because its hard to imagine how it will transform our society. I want to see a Star Trek economy/society in my lifetime. I only life once. Btw. AI/LLM/Machine learning is the gateway technology for robotics, this will affect even more.

It's the wrong way around. If we get AGI (or any well working AI) before we abandon capitalism it's going to be a huge disaster. A handful of even richer even more powerful very greedy people will have all the wealth and everyone else will have nothing. I mean, there was a WW3 in Star Trek, so maybe it was that path that humanity took in Star Trek anyway?

Thankfully there's nothing to suggest that LLMs will ever bring us AGI (or any well working AI)

Re: GPT-2: Too Dangerous To Release (2019)

#113

Post by esafak »

GTP wrote: post #92
Qhemlomo wrote: post #64

Earlier quoted context omitted.

You know what the problem is in software engineering? A LOT of people have no clue what good software engineering is. I was working in a company before which used md5 in 2015! Databases on the internet with a 5 character password. No tests. A person i know would have broken the whole production DB if i wouldn't have stoped the PR. Another ex-collegue thought its okay to 'encrypt' with a basic shift cyper creditcard d…

> You know what the problem is in software engineering? A LOT of people have no clue what good software engineering is. Indeed. Is Mythos going to change this?

Yes. It is going to make better decisions for people that don't know better.

Re: GPT-2: Too Dangerous To Release (2019)

#114

Post by nonethewiser »

watwut wrote: post #99
nonethewiser wrote: post #69

Earlier quoted context omitted.

> The president of the United States tweeted an AI generated image of himself as Jesus Christ descending from the sky and saving a sick person. Does that say anything about AI or everything about Donald Trump?

Both. And even more for people who still defend republican party ... which stands behind and supports Trump 100%

But what does it say about AI?

That it can depict Trump as Jesus? You don’t need AI for that.

Re: GPT-2: Too Dangerous To Release (2019)

#115

Post by protocolture »

oathvz wrote: post #22
minimaxir wrote: post #8

Back in 2019, it was more fair to have caution around the larger GPT-2 models since robust text generation (by 2019 standards) was a complete unknown. For something like Mythos in 2026, where now the social implications of better LLMs are more understood, it's more fair to call it (EDIT: specifically, the declaration of its danger) a marketing gimmick.

This is a natural follow up question -- what kind of an escalation or message should frontier labs/companies publish to be seen as genuine and not marketing gimmick?

I dont think they can, their rhetoric is so silly these days that if they did accidentally release something dangerous no one would believe them. The Boy who cried Rokos Basilisk I guess.

Re: GPT-2: Too Dangerous To Release (2019)

#117

Post by gnunez »

I remember those days. They released smaller versions of the model that were interesting but pretty useless. The wild part is, about 5 years later I was able to train the same model from scratch, using a couple of gpus in my home office, following Kaparthy’s online lectures, for a fraction of the cost. Long Live Open Source!

Re: GPT-2: Too Dangerous To Release (2019)

#118

Post by realusername »

jason_oster wrote: post #98
realusername wrote: post #83

Earlier quoted context omitted.

I'd say creating a project is 5% of the job and maintaining it over time 95% of it. It's true that they can start amazing projects without guidance but then the real work begins.

There is almost no maintenance work for bespoke apps apart from infrequent updates to keep OS and hardware compatibility as the environment slowly changes. Keep in mind, these are not products in the endless feature treadmill promoted by scrum.

I've worked on an app and I'd say the opposite, I even abandonned the app at some point due to the maintenance work involved.

Re: GPT-2: Too Dangerous To Release (2019)

#119

Post by Qhemlomo »

realusername wrote: post #83
Qhemlomo wrote: post #38

Earlier quoted context omitted.

My partner wrote an android app which was doing what she wanted to do. She did this experiment 5 month ago and she did this in one day. My wife has 0 knowledge how any of this works. That was shocking to see. Progress is not stoping and Fable proves that.

I'd say creating a project is 5% of the job and maintaining it over time 95% of it. It's true that they can start amazing projects without guidance but then the real work begins.

Thats what the agentic layer will fix, but this is currently getting build.

Re: GPT-2: Too Dangerous To Release (2019)

#120

Post by Qhemlomo »

GTP wrote: post #92
Qhemlomo wrote: post #64

Earlier quoted context omitted.

You know what the problem is in software engineering? A LOT of people have no clue what good software engineering is. I was working in a company before which used md5 in 2015! Databases on the internet with a 5 character password. No tests. A person i know would have broken the whole production DB if i wouldn't have stoped the PR. Another ex-collegue thought its okay to 'encrypt' with a basic shift cyper creditcard d…

> You know what the problem is in software engineering? A LOT of people have no clue what good software engineering is. Indeed. Is Mythos going to change this?

On one side, it means that a certain amount of business will just use it even if you think its not safe/good enough and they will throw out people and will still succeed.

And on the other side: yes because they will also use LLM review or other tooling and will be fine whatever the 'security llm agent' tells them.

Post reply on HN