This comment isn't related to Paris but I just wanted to tell Gary, if he didn't watch, that Salesforce's superbowl AI ads were hilarious and embarrassing. The first one was Matthew trying to argue that without AI people wouldn't be able to find their airport terminal - as if people haven't been doing this successfully for decades. The second was Matthew trying to argue that absent AI a restaurant would seat you outside in pouring rain and give you food you didn't order lol. They are really trying to make us think we are helpless. I'm hoping that people know better - I worry when I hear that school's are integrating this into their curriculum. It makes me think of how cabbies used to be skilled operators who knew the roads and traffic patterns in their city. Now they are clueless operators who follow the directions.
The only way I see the current gen of AI succeeding is if the overlords succeed in training us to be helpless.
Why did they bother having a summit? It's literally an AI meeting - so why not just send in the various avatars and have them interface with each other and produce a well-polished knock-off of something-something-agreement? #WeAreTheBorg
Truth. For practical real world applications that can actually make a difference LLM is just about the worst. I've worked with Speech to Text and Text Extraction where a level of confidence is assigned to results, and those are WAY more cost effective and useful than all the tinkering and IP theft that's been done with LLMs.
Even then you still need a human to go over the outputs that get flagged as uncertain. It just speeds up things like medical transcription and data entry, and removes some of the misery from tasks like processing invoices.
According to Judd Legum recent Substack article, there is now a ban on certain words within the US Government, one of which is 'inclusive.' It would seem pretty likely, just for starters, that the same government implementing this ban is not about to sign onto a diplomatic declaration that contains it. Would a simple rewording solve the matter? Unlikely, given there is no there, there.
Sustainable and inclusive? What does that even mean? Deepseek comes out with truly innovative technology with R1 and everyone is concerned that it won't talk about Tiananmen Square or the Uighurs. Really? American LLMs have their own sacred cows, so what's the big deal? As we have seen with other agreements like this, the West is the only group of nations that even tries to follow the restrictive platitudes, so why tie our hands with imaginary handcuffs? Just duke it out in the market!
I presume the value in such gatherings comes from (i) work done in advance (not much it seems), and (ii) the side meetings, rather than the main event itself.
Not surprising. The likely outcome in general of such global initiatives is to simply move money around between powerful hands all while the reality of the crisis becomes less discernable through the fog of political battles.
How convenient that the UK happens to be 1 of the few European countries that is not part of the EU and therefore not under the jurisdiction of the GDPR. Looks like the UK seems to believe that selling out their people and country to tech bros and be part of the AI race will somehow restore the UK's status after the disaster that was Brexit. Given the current and foreseeable state of affairs, marginal gains and slowing progress and new competitors like China and open source startups, I suspect they will all be quite disappointed.
This is like boilerplate industry talking points - why? Such a missed opportunity. So many technologies, societal impacts, questions, harms, challenges, and needs - and that's is the statement?
I love this! And the fact is that AI is not actually a threat anytime soon, so everything is actually fairly fine (for now). Why? Because... dumbass.
There is no dumbass like an academic dumbass who will go to his death pimping for the AI equivalent of the phlogiston theory of heat.
It is similar to the situation in biological warfare. Despite everything the brilliant worthies at the NSABB can do to disseminate and instruct enemies who want to kill us, it doesn't work. Why? Because... dumbass. First, thankfully, the NSABB's ideas of what is a problem are fairly disjoint from what is, so far. Second, said enemies are mentally hamstrung by their fundamentalist straitjacket that can't reason its way out of a wet paper bag---although the West is trying hard to catch up by raising children on a steady diet of word-salads and race-to-the-bottom education standards. (Yes, if we get to the end of the 21st century without a major bioweapon plague, I will be astonished; but, at present, I see what makes things not go that direction right now. It ain't some clever department at CIA preventing anything. No, it's dumbass all around! Dumbass for the win! Yay, dumbass!)
Here's a paper on how dumbass works in biodefense.
“... we only became aware of them (biological weapons), when the enemy drew our attention to them by repeatedly expressing concerns that they can be produced simply with easily available materials....” -al Zawahiri internal memorandum, April 15, 1999
This comment isn't related to Paris but I just wanted to tell Gary, if he didn't watch, that Salesforce's superbowl AI ads were hilarious and embarrassing. The first one was Matthew trying to argue that without AI people wouldn't be able to find their airport terminal - as if people haven't been doing this successfully for decades. The second was Matthew trying to argue that absent AI a restaurant would seat you outside in pouring rain and give you food you didn't order lol. They are really trying to make us think we are helpless. I'm hoping that people know better - I worry when I hear that school's are integrating this into their curriculum. It makes me think of how cabbies used to be skilled operators who knew the roads and traffic patterns in their city. Now they are clueless operators who follow the directions.
The only way I see the current gen of AI succeeding is if the overlords succeed in training us to be helpless.
Yeah, that was just comedy all the way through.
Why did they bother having a summit? It's literally an AI meeting - so why not just send in the various avatars and have them interface with each other and produce a well-polished knock-off of something-something-agreement? #WeAreTheBorg
This is clearly a summit to bless the objectives of LLM based AI. Do the attendees even recognise that there are dozens of different AI technologies?
There doesn't seem to be anything about the problems of uncontrollability and the blackbox nature of most AI Technologies.
Truth. For practical real world applications that can actually make a difference LLM is just about the worst. I've worked with Speech to Text and Text Extraction where a level of confidence is assigned to results, and those are WAY more cost effective and useful than all the tinkering and IP theft that's been done with LLMs.
Even then you still need a human to go over the outputs that get flagged as uncertain. It just speeds up things like medical transcription and data entry, and removes some of the misery from tasks like processing invoices.
Looks from the snippet to be a declaration written by ChatGPT...
According to Judd Legum recent Substack article, there is now a ban on certain words within the US Government, one of which is 'inclusive.' It would seem pretty likely, just for starters, that the same government implementing this ban is not about to sign onto a diplomatic declaration that contains it. Would a simple rewording solve the matter? Unlikely, given there is no there, there.
There, there. I no too much. We no more.
Gary no's best!
Sustainable and inclusive? What does that even mean? Deepseek comes out with truly innovative technology with R1 and everyone is concerned that it won't talk about Tiananmen Square or the Uighurs. Really? American LLMs have their own sacred cows, so what's the big deal? As we have seen with other agreements like this, the West is the only group of nations that even tries to follow the restrictive platitudes, so why tie our hands with imaginary handcuffs? Just duke it out in the market!
I presume the value in such gatherings comes from (i) work done in advance (not much it seems), and (ii) the side meetings, rather than the main event itself.
Not surprising. The likely outcome in general of such global initiatives is to simply move money around between powerful hands all while the reality of the crisis becomes less discernable through the fog of political battles.
"US and UK, two of the most important countries in AI industry", a joke that just keeps on giving. 😆
How convenient that the UK happens to be 1 of the few European countries that is not part of the EU and therefore not under the jurisdiction of the GDPR. Looks like the UK seems to believe that selling out their people and country to tech bros and be part of the AI race will somehow restore the UK's status after the disaster that was Brexit. Given the current and foreseeable state of affairs, marginal gains and slowing progress and new competitors like China and open source startups, I suspect they will all be quite disappointed.
This is like boilerplate industry talking points - why? Such a missed opportunity. So many technologies, societal impacts, questions, harms, challenges, and needs - and that's is the statement?
If anyone actually wants to regulate anything they'd have to give up money, and OF COURSE that didn't happen
I love this! And the fact is that AI is not actually a threat anytime soon, so everything is actually fairly fine (for now). Why? Because... dumbass.
There is no dumbass like an academic dumbass who will go to his death pimping for the AI equivalent of the phlogiston theory of heat.
It is similar to the situation in biological warfare. Despite everything the brilliant worthies at the NSABB can do to disseminate and instruct enemies who want to kill us, it doesn't work. Why? Because... dumbass. First, thankfully, the NSABB's ideas of what is a problem are fairly disjoint from what is, so far. Second, said enemies are mentally hamstrung by their fundamentalist straitjacket that can't reason its way out of a wet paper bag---although the West is trying hard to catch up by raising children on a steady diet of word-salads and race-to-the-bottom education standards. (Yes, if we get to the end of the 21st century without a major bioweapon plague, I will be astonished; but, at present, I see what makes things not go that direction right now. It ain't some clever department at CIA preventing anything. No, it's dumbass all around! Dumbass for the win! Yay, dumbass!)
Here's a paper on how dumbass works in biodefense.
https://www.omicsonline.org/security-in-a-goldfish-bowl-the-nsabbs-exacerbation-of-the-bioterrorism-threat-2157-2526.S3-013.php?aid=11953
“... we only became aware of them (biological weapons), when the enemy drew our attention to them by repeatedly expressing concerns that they can be produced simply with easily available materials....” -al Zawahiri internal memorandum, April 15, 1999
AI is beyond the control of any one company or on country, and those who have power in both areas are going crazy as a result. https://www.danablankenhorn.com/2025/02/revenge-from-the-bottom-up.html
There won't be any consensus or regulation on AI for the foreseeable future. Also, Europe does nothing except create rules.
This is what we should have been talking about https://www.stuckpodcast.com/p/how-to-measure-ethical-ai-a-framework?r=53raj