GPT-2: Too Dangerous To Release (2019) (2022)
GPT-2: Too Dangerous To Release (2019)
11–20 of 139 posts
Re: GPT-2: Too Dangerous To Release (2019)
#12Seven years of this insufferable brand of "Oh it's so dangerous, I sure hope no one gives us a ton of money and takes us seriously" marketing and people are still falling for it at scale.
Every night I am wracked by grief and anxiety that we might deliver too much value to our investors and shareholders. If only someone would create legislation that would mildly inconvenience us while crippling potential competitors!
Re: GPT-2: Too Dangerous To Release (2019)
#13Unfortunately, Anthropic and Claude models have joined the ranks of Mind-killer topics where the signal to noise ratio in any discussion has dropped through the basement.
Re: GPT-2: Too Dangerous To Release (2019)
#14Seven years of this insufferable brand of "Oh it's so dangerous, I sure hope no one gives us a ton of money and takes us seriously" marketing and people are still falling for it at scale.
They feared that GPT-2 could break all Spam filters.
And tbh do you prefer companies not taking anything serious?
Opus 4.5 def changed a lot already, GenAI changed a lot.
Certain jobs are gone. Do you think the person who was translating text doesn't deserve to be taken serious?
I haven't written code in a few month now and the quality of these coding agents is not getting worse, they are getting better.
All of this is transformable and we just started. GPT-3 came out in 2020 and public got access to it only 2022.
The last 4 years do not feal like 4 years and we are still progressing.
We have to also ask us as a society what is happening to young people. Even if we accept that we still hire juniors, they themselves have to completly rethink how they learn and how they work.
Re: GPT-2: Too Dangerous To Release (2019)
#15they knew the risks and went ahead anyways, making them LIABLE for the damages that followed.
Re: GPT-2: Too Dangerous To Release (2019)
#16I believe it was a marketing strategy.
Re: GPT-2: Too Dangerous To Release (2019)
#17In hindsight, they were entirely correct. The social damage caused by low cost content generation that’s hard to distinguish from human authorship is astronomical. You don’t need to entertain the more ridiculous doomsday scenarios to wish that this technology had never been created.
Of course, this damage could still be enabled with just hosted access to the models, restricting access to the model files themselves did not stop that
Re: GPT-2: Too Dangerous To Release (2019)
#18Say hypothetically that they were concerned that GPT models would see widespread abuse, for example by students cheating on homework assignments, in a way that could cause likely-irreversible societal changes some of which are harmful. Can we confidently say they were wrong?
The dangerous use cases back in 2019 were spam and phishing and GPT-2 1.5B was nowhere near good enough to do homework assignments. No one envisioned how LLMs would develop.
Re: GPT-2: Too Dangerous To Release (2019)
#19In hindsight, they were entirely correct. The social damage caused by low cost content generation that’s hard to distinguish from human authorship is astronomical. You don’t need to entertain the more ridiculous doomsday scenarios to wish that this technology had never been created.
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.
Re: GPT-2: Too Dangerous To Release (2019)
#20In hindsight, they were entirely correct. The social damage caused by low cost content generation that’s hard to distinguish from human authorship is astronomical. You don’t need to entertain the more ridiculous doomsday scenarios to wish that this technology had never been created.
I certainly cannot survive much more of the AI memes generated about our so-called Commander in Chief with a fake bodybuilder mystique... you are absolutely correct, this kind of material is psychologically damaging. And a huge distraction from the genocide by the "best friend and ally" of the US. Heart wrenching and extremely damaging hasbara - just please stop, haven't you stolen and killed enough guys? This is _not_ the old American West when communications were few and it was most often a tale of solitary survival. It's organized Nazi-esque kill, command and control, enable by so-called AI to take some guilt off the shoulders of those pushing buttons and pulling triggers.