00:00 / 00:00
Sleep timer ended
⏪ 10s
⏩ 10s
1.2K views
1 month ago
Artificial intelligence is advancing fast, and so are concerns about where it's heading. Some of the world's biggest AI companies says the want to join forces t…
...more
18 Comments
The sandbox was faulty and they have admitted so. The break could easily be tested again with the same scenario and better logging. Yet that would take money just to see why it broke out which is basically a bad sandbox. It’s very likely the frontier companies wanting to remain frontier as long as possible.
LLMs have a great future. Unlike politicians, they actually solve real-world problems. As a starting point, you could enforce data laws. There are other imminent matters such as espionage. 🎉
according to my calculations, the odds of the AI "killing all humans by 2036" is about 1:220,265 but I failed statistics
Not sure. But misuse will be a nightmare
Well, once half the world is unemployed you’d wish you could be back.
If you tell the affected "AI" that it will be destroyed, if it is programmed to preserve itself at all costs, that's where the fun begins.
You now you can always pull the plug at the datacenters running the „rogue“ AIs.
"Moreover, this will occur in tandem with the equally steadily rising and globally spreading irrationality and depravity of the Earth's population, who will, in a completely thoughtless and utterly irrational manner, submit themselves to an emerging artificial intelligence and become entirely dependent on it, just as they are on many new types of technical products, such as various remote transmission technologies, communication technologies, information technologies and image transmission technologies, etc."<br><br>Source: Billy Meier contact report 928 from the future of mankind website.
Seriously 😒 <br>What are you talking about over regulations!<br>There is no regulation at all 🤦<br>It's the reason for companies wanting regulations.
Al for National security is largely governed by classified executive directives, not public regulation. Much of the most advanced Al work in defense and intelligence is classified, meaning it receives no public safety review, no transparency requirements, and no independent audit. The companies doing it are the same companies asking for safety regulation of everyone else. If these firms carry trillion-dollar obligations, they cannot allow a commodity model where inference approaches marginal cost. They need pricing power.<br>They need tiered access. They need a regulated moat that raises fixed costs for competitors while they absorb those costs through scale. Safety regulation is the most politically palatable form of moat-building because it sounds like public protection.
The last thing we want is the government as the umpire.
This is the most coherent treatment of this issue I've seen so far. Altman and Amodei each have a string of multi-million-dollar investors who were promised a product of atom bomb-level . But performance of both frontier models seems to be leveling off and, arguably, regressing to the mean. It will be much easier to explain model slow downs to those investors if Altman and Amodei can blame it on necessary government control. Just a year ago they were lobbying the President for an EO that would grant them immunity from state-level regulation (something that Thomas Jefferson would never have imagined a Republican considering), and now...
I think we have missed the point. These companies are responsible to shareholders, so they cannot restrain themselves individually and voluntarily, because this will simply hand any development advantages they may hold to their competitors. Therefore, regulation has to be wide-ranging and unilateral, so that all competing companies will be restrained TOGETHER. I think these CEOs have genuinely recognised this, and this is why regulation must come from government, and, more importantly, on an INTERNATIONAL basis. THEY ARE RIGHT!
Literally no one wants it, but they are pushing it into our lives. Why.
High IQ yes... (high Idiotic Qualities)
I'm sick of everything AI.
AI isn't going to kill anyone. People will kill other people with autonomous weapons.
If AI kills any one person the ceo and staff at the company should go to jail for murder
Up next
From this channel
