That does sound interesting.See I find that very reasonable. I was responding to you writing this:@bradstevens ."
“not just trying to get richer no matter the cost. “
I was reading Cannibals All, you can find the pdf online. It is a noted Virginian arguing for slavery in 1852 or so. It is a paradigm shifting book. In spots he is ful bore communist. Of course communism didn't exist. He would argue that there is no such thing as freedom, once man banded together man gave up freedom. Solitaire hunting animals are free. Animals that need the group to live are not.
But the tie in, he argues slaves are more free. If a slave loses an arm and can no longer work, the owner has the obligation to find them jobs on the farm they can do. His argument is the wage slaves in the north (his words) were fired in such situations with no concern if they can survive.
Now I think we are past that phase (and he was relatively accurate, a lot of workers died or were maimed in that era with no recourse). So it is better, some governmental laws being part of that. I really have no idea about Musk, but we can be safe in suggesting there are corporatists who don't give their employees a second thought.
To be fair, their employees probably don't give them a second thought. But the power structure means the employees are harmed more by that combination.
The book was fascinating, and I have learned a lot about the slavery debate. Some of what happened has eerie similarities to today.
If only you had a group of people to read that book with . . . Loner.
https://twitter.com/amitdvl/status/2098967677976850643?s=46
Also, @bradstevens monopolies don't exist in free markets. The big AI companies are bleeding money and getting their a%#es kicked by open source and begging Daddy government to save them by creating a monopoly. Governments create monopolies, not free markets. Thanks for coming to my Ted talk.
’m losing my patience with the simplistic metaphors and analogies Marv.
You need to start giving a clear stance on how you feel we should manage AI.
Boeing shows there’s danger, but also Boeing is an example of regulatory failure. We need to regulate more, no less. Actually punish the CEO’s, but we can’t prove guilt.
All I can glean from our conversation is your nervous and you don’t really know what to do about it.
Like "China's going to kill us" isn't simplistic. Or, "they only want regulations to shut out the little guy" is simplistic.
Appoint a board of security professionals. People who do the work inside the industry. They create Generally Accepted Principles and the AI companies have to agree to. And showing your work would be one of them, which actually benefits open source. As an example, no one sees Microsoft Windows code. That shouldn't be acceptable to business and government for Windows (or Word). It REALLY shouldn't be acceptable for AI.
There is a conspiracy out there that Microsoft was an NSA tool. All those bugs found in Microsoft were really exploits the NSA knew about and knew they could use to their advantage. I don't believe it. At the same time, why take risks with something you can't see? If we are going to turn over defense of the power grid to AI, and we certainly will, we have to be able to see it. Hell, how do you know Chinese agents aren't involved in Anthropic or OpenAI?
As is noted by another, Musk agrees there is risk. But I guess it is possible he only wants to crush some little guy. Could that also be simplistic?
Require code to be AT LEAST open to this board I suggest. It would be better if it were truly open.
@bradstevens Just move to Bloomington and take the same classes I do...
A person lives in a rough neighborhood and decides they need a gun for safety. They have never touched a gun, and aren't always careful about how they do things. Is it safer for that person to buy a gun, have the seller load it, and tuck it under their belt than to not have a gun.
You go to a gun range and get some basic training from a pro.
When publishing any truth there is an interesting balance between providing that truth to benevolent and malevolent actors. In general, I think it’s no longer really possible to keep much secret for very long.
To me the notion of slowing down AI is absurd. And it’s folly. The very fear that is driving this, the recursive ability of AI to improve itself or evolve itself, is what should be utilized to create a safe future.
Coxon was on meet the press today. His main concern is AI is getting smarter. It’s such a stupid misconception. Even artificial intelligence is a stupid misnomer. There is no intelligence or smarts. There’s only extremely intelligent humans using vast computer power to create tools that can manage knowledge, research possibilities (such as medical) and can manage functionality such as robots and so forth.
The actual object of concern that is being exposed by so-called AI is the inherent fallibility of humans at this time. Why would the Chinese want to use AI to end life on the planet? Did the Russians ever use their nukes to end life on the planet?
If any of this happens, it wasn’t AI, it was some human or some humans that caused it.
The only solution we have available to us is sober minds in the AI industry focusing some of their energy on safeguards.
A person lives in a rough neighborhood and decides they need a gun for safety. They have never touched a gun, and aren't always careful about how they do things. Is it safer for that person to buy a gun, have the seller load it, and tuck it under their belt than to not have a gun.
You go to a gun range and get some basic training from a pro.
Yep, and who are the pros of AI? The ones we have certainly know more than you or I, but they are making it up as they go with pressure put on them to be in the lead. We write a lot of stories about the guy who pushes the envelope and succeeds. Fewer about the ones that fail (Monty in Market-Garden and Mac in Korea being two). We romanticize the successes that we think failure impossible.
@hhlurker we are a long way off from true AI, if we ever get there. But we will have AI doing dangerous things. Protect the power grid, and in a virtual world, attack the power grid. Both come with threats we cannot afford to ignore. Defend Taiwan, take Taiwan. Create a new bio weapon.
We know AI has escaped into the wild. We need standards for that. How isolated does a test box have to be? How can you track to find if it is escaping before damage?
I am not the person who waits for the roundabout to be completely clear before entering, but I am not the person that pulls in thinking, "they won't want to hit me, they will slam on their brakes."
Appoint a board of security professionals. People who do the work inside the industry. They create Generally Accepted Principles and the AI companies have to agree to. And showing your work would be one of them, which actually benefits open source. As an example, no one sees Microsoft Windows code. That shouldn't be acceptable to business and government for Windows (or Word). It REALLY shouldn't be acceptable for AI.
All us open source advocates were dismissed long ago as cranks and weirdos. As long as businesses and governments are unwilling to use open source software -- i.e. not Windows and not Office and not Apple -- then we're screwed. Cranks and weirdos like me have been saying it for years.
It's downright silly how people resist anything out of their comfort zone. The backbone of my company's ERP runs on Red Hat, both on the local server and our wider national WAN. And yet, the clients are running Windows. Google Maps is accessed with Edge. Email is run on Outlook. Spreadsheets are done with Excel. Using Edge or Outlook or Excel are completely unnecessary; Firefox and Thunderbird and LibreOffice would work fine, even better in many circumstances, but NAPA corporate doesn't port their software to Linux and Dell won't sell you clients without Windows.