The genie escapes: Stanford copies the ChatGPT AI for less than $600

submitted by hfxB0oyA from (newatlas.com)

all 10 comments
sorted by:
top
newinsightfulfun

[–]HiddenFox 3 insightful - 1 fun3 insightful - 0 fun4 insightful - 1 fun -  (2 children)

Like all things AI, the path this technology takes will be utterly unpredictable. Given this article to be true, I never would have guessed you could reproduce ChatGPT level AIs at that cost and with this ease.

I do however think there is a silver lining (or possible disaster) to this. At some point with this technology running muck it will be impossible to believe anything you see, hear or read on the internet. It will render it useless for information purposes. Only walled off ecosystems will become trustworthy. But that too will also lead to manipulation as we have already seen. (IE google censorship.)

It will take some major event in deception where so many are fooled into something so preposterous, but in the end people's attitude will shift to the internet. Being nothing more then an endless space of lies and misinformation people will have little choice but to go backwards...

[–]Canbot 1 insightful - 1 fun1 insightful - 0 fun2 insightful - 1 fun -  (1 child)

The wealthy and powerful already had that type of technology, if slightly different. They simply employed armies of shills with force multiplier tech that allowed each shill to control thousands of accounts and personas.

You see it the most during election season. Shareblue, acorn, JIFD, etc. It is what allows them to build fake "grass roots" movements that (((they))) use to attack and manipulate the public and the government with. Groups like Antifa and BLM.

[–]EddieC 1 insightful - 1 fun1 insightful - 0 fun2 insightful - 1 fun -  (0 children)

You see it the most during election season. Shareblue, acorn, JIFD, etc. It is what allows them to build fake "grass roots" movements that (((they))) use to attack and manipulate the public and the government with. Groups like Antifa and BLM.

So, do you see this (#1-#3) then?
 

[–]EternalSunset 2 insightful - 1 fun2 insightful - 0 fun3 insightful - 1 fun -  (0 children)

It also can run locally on your own hardware btw. I have seem users who say it runs fine on a RTX 3080.

[–]dicknipples 2 insightful - 1 fun2 insightful - 0 fun3 insightful - 1 fun -  (0 children)

Sure - $600 + the time of a team of Stanford researchers (a lot of salaries and student loans); but it is great that it's open source. As noted in the article, other companies are developing similar AI, which will soon be as ubiquitous as search engines.

[–]SoCo 2 insightful - 1 fun2 insightful - 0 fun3 insightful - 1 fun -  (0 children)

I'm not sure why everyone is pretending ChatGPT is something new. It is just hard and expensive to make at that level, as the article mentions, so most companies who did spend the effort, sold its use to those who used it for oppression and weaponized anti-democratic war, anti-civil rights, or election campaigns for the last 10+ years.

[–]package 1 insightful - 1 fun1 insightful - 0 fun2 insightful - 1 fun -  (1 child)

These AI are impressive, but the real breakthrough will come with rule extraction. Knowing how to generalize the abstract concepts these models are successfully mastering would open the door to extremely accurate and complex behaviors. It might even lead to deciphering and possibly augmenting the human brain.

[–]iamonlyoneman 2 insightful - 1 fun2 insightful - 0 fun3 insightful - 1 fun -  (0 children)

There was a documentary about this, actually

[–]William_World 1 insightful - 1 fun1 insightful - 0 fun2 insightful - 1 fun -  (0 children)

I have a feeling you're all AIs on this site

[–]Canbot 1 insightful - 1 fun1 insightful - 0 fun2 insightful - 1 fun -  (0 children)

While the text generation may be copied, the real power comes from the massive data sets behind it. I have my doubts about it.