I watched the video. A lot of people are expressing these concerns. But who is going to bang heads together on it?
From my POV, using Claude as a “pretty good software engineer” as well as a “super google”, it is a fantastic productivity tool if you are competent. So e.g. in software you have to remain the “architect”, as well as knowing enough to understand what it has done. Then, your productivity can go up 10x or more.
People who are not competent will not be able to use it effectively and will just keep getting into a mess. There are many cases of corporate users blowing away 5 digits a year and getting junk – by letting idiot users (a big % of any large company, because a small company can’t employ useless people) loose on it, and running it in a loop.
I am worried about the deeper implications. AI is an amplifier of human intention…good or bad. If we feed it bad, it can be the baddest of bads, to the point that we have no control over it…
“We” will always control it. It is only a computer after all. Only when some dick does what they do in movies and connects some “AI” machine to an ICBM launch button will there be real issues.
The internet has always been an indicator of the lowest common denominator of human nature.
What is “fun” for sure is “AI” finding new back doors in server code etc in common use. That’s sure to happen.
I would seriously worry if I was young, and looking for a job as a less than very very good software engineer 
There are always people who for various reasons do not have a grip on their life, and they will start having conversations with “AI”. I actually know of some.
Those Chinese running robots (current news) are amazing. But, have you heard about the one which got disqualified for testing positive for WD40?
Peter wrote:
Only when some dick does what they do in movies and connects some “AI” machine to an ICBM launch button will there be real issues.
For sure the military wants to do this. In the early 1980s the USA wanted to station nuclear armed Pershing ballistic missiles in West Germany. There was a big fuss as the flight time from the “iron curtain” was so short that, arguably, manual launch decision would be too slow making automated procedures necessary.
I recall attending the International Joint Conference on Artificial Intelligence, at that time the major world conference on AI, in Germany (Karlsruhe), 1983, as a Ph.D. student. Pershing missiles was the big issue, with panel discussions and demonstrations outside and inside the conference venue. I also recall the grand old man of AI – John McCarthy, being obviously not amused by the commotion.
For sure the military wants to do this. In the early 1980s the USA wanted to station nuclear armed Pershing ballistic missiles in West Germany. There was a big fuss as the flight time from the “iron curtain” was so short that, arguably, manual launch decision would be too slow making automated procedures necessary.
I am 100% sure that is total fiction. None of the detection systems are 100% reliable. They used, and still use, initially, the IR launch signature, which will never be 100%.
Pershing missiles was the big issue, with panel discussions and demonstrations outside and inside the conference venue.
Same here in the UK. Greenham Common became a well known women’s camp
There was real doubt whether the missiles would actually work, because they used terrain following radar, and the USSR landscape could change a lot in heavy snow. But the Russkies could not be sure either way, which was all a part of the deterrent.
Peter wrote:
There was real doubt whether the missiles would actually work, because they used terrain following radar, and the USSR landscape could change a lot in heavy snow. But the Russkies could not be sure either way, which was all a part of the deterrent.
Are we talking about the same thing? Pershings were ballistic missiles, not cruise missiles.
OK; different. I googled it: Ninety-six BGM-109G Gryphon Ground-Launched Cruise Missiles (GLCMs) were scheduled to be based at RAF Greenham Common. These were the ones with terrain following guidance.
But same “anti US political protest” stuff.
And no way could I believe anybody would make launching of nukes (or anything else actually, because big conflicts can start with small provocations) automated.
If anything, I think the effect of “AI” on our large society is being overestimated. There is a fantastic amount of marketing BS going around; every company is claiming to use “AI”. I am sure the firm emptying the dustbins would use it if they discovered it. It will transform tasks like software. It will transform repetitive tasks. It is even coming up with theorem proofs which have eluded mathematicians for years (although opinion differs on this).
Probably the biggest effect will be in medicine, where we expect 100% competence but so often do not get anything like it. Like in software, the workers who are less than very good will get pushed out.
It’s great if you run your own tech business because you can use it to great advantage, and build up the company with a lot fewer people, and that’s always a good thing
Are you not worried about more insidious ways than controlling an ICBM?
Like your social media, your digital identity, your bank account or anything SW-controlled that has some relevant bearing in your life..?
Can you give specific examples?
I would hope the bank account login is safe, otherwise we are all in a big mess Anybody can set up a fake FB account which is a copy of mine or yours, and this happens all the time. You don’t need “AI” for that.
I don’t think I have a digital identity.
The biggest problem is hacking into systems which are important, and “AI” will certainly assist with that. But one can also use it to examine and harden the same systems. After the recent DDOS attack on EuroGA (which incidentally is still being continuously attempted) this was in fact done.
Chloe’s video explained how a S/W review by Claude had uncovered tens of thousands of security vulnerabilities which human programmers had not detected.
If AI is used to develop S/W and, by training, Claude learns to deliberately hide some of them to include a Trojan or some similar functionality in certain critical S/W in order to allow undue external control of some of its functions without its creators knowing…
