Air Force colonel 'misspoke' when he said an AI-drone 'killed' its human operator
- Reference: 1685961428
- News link: https://www.theregister.co.uk/2023/06/05/us_air_force_colonel_admits/
- Source link:
Colonel Tucker "Cinco" Hamilton reportedly described the possibility of the AI system turning against its human handler – attacking and killing the operator in a computational simulation – as a cautionary tale. The Royal Aeronautical Society, which hosted the Future Combat Air & Space Capabilities Summit, quoted Hamilton in a report that was quickly picked up by the press (including [1]El Reg ).
Ann Stefanek, a spokesperson from the US Air Force, however, told The Register : "The Department of the Air Force has not conducted any such AI-drone simulations and remains committed to ethical and responsible use of AI technology. This was a hypothetical thought experiment, not a simulation."
[2]
"It appears the colonel's comments were taken out of context and were meant to be anecdotal," she said. A representative from The Royal Aeronautical Society, on the other hand, said it was Hamilton who made the errors.
[3]
[4]
In a statement to the Society's magazine Aerospace , the representative said the colonel admitted he "misspoke" and that he was describing a hypothetical thought experiment and not one that had actually taken place at the US Air Force.
"We've never run that experiment, nor would we need to in order to realize that this is a plausible outcome," Hamilton is quoted as saying. "Despite this being a hypothetical example, this illustrates the real-world challenges posed by AI-powered capability and is why the Air Force is committed to the ethical development of AI," he said.
Getty files second UK lawsuit against Stability AI
Getty called judges at the High Court of Justice in London for injunctive relief to stop Stability AI selling its text-to-image software in its second lawsuit against the company.
The stock photo supplier [5]sued the startup in January this year, and accused it of violating copyright laws by scraping the company's images to train its Stable Diffusion models. In a second filing [6]reported this week by Reuters, Getty has gone one step further and asked judges to block Stability from selling its AI tools altogether.
[7]
It has reportedly asked the High Court for Stability to destroy all Stable Diffusion models that were allegedly unlawfully trained on its images or hand over the software, and is suing for damages. The amount has not been finalized.
Meanwhile, Stability AI has not yet filed a defence to fend off Getty's claims and faces a separate, additional lawsuit filed by Getty in US courts.
You can blame AI for some layoffs, says report
A report from outplacement firm Challenger, Gray & Christmas suggests that about 3,900 jobs have been cut and replaced by AI in May alone.
The latest [8]Challenger report showed just how frosty the labor market is right now. US-based employers axed 80,089 jobs in May – a 20 percent increase compared to the previous month and a 287 percent jump from the same month last year.
"Consumer confidence is down to a six-month low and job openings are flattening," Andrew Challenger, the company's senior veep, explained in a statement. "Companies appear to be putting the brakes on hiring in anticipation of a slowdown."
[9]
Table four in the report reveals that the top reasons for layoffs are the current tough global economic climate still reeling from the COVID-19 pandemic and the ongoing Russia-Ukraine war affecting supply chains. In tenth place is AI. Challenger, Gray & Christmas reckon nearly 4,000 jobs have been replaced with the automated technology.
Lured by the idea of increased productivity and efficiency, businesses from all sorts of industries are rushing to incorporate AI. Some, like IBM and BT, have [10]previously said that generative AI will replace thousands of workers in HR and customer service roles.
It may not be all doom and gloom, however. Other big orgs like JPMorgan have advertised 3,651 AI-related jobs from February through April – more than any other bank, [11]according to Bloomberg.
Microsoft to work with third party cloud provider for extra GPUs
Microsoft has reportedly signed a deal with cloud computing startup CoreWeave worth potentially billions of dollars for access to its GPUs, in a bid to help ChatGPT keep pace with demand.
Sometimes ChatGPT cannot process user requests since the service is often overwhelmed. Training and running such AI models is computationally intensive, and requires huge amounts of hardware. Even Microsoft's own Azure cloud platform cannot always support OpenAI.
[12]Foxconn is ecstatic you're all going gaga for AI servers
[13]Microsoft Windows latest: Cortana app out, adverts in
[14]Amazon finds something else AI can supposedly do well: Spotting damaged goods
[15]Amazon finds something else AI can supposedly do well: Spotting damaged goods
That's why Redmond has invested in CoreWeave – a startup that has amassed tons of GPUs that were previously used to mine cryptocurrencies, [16]according to CNBC. CoreWeave is now a cloud provider, and focused on renting out its hardware resources.
The startup [17]said last week it had secured $200 million in its latest round of funding, led by asset manager firm Magnetar Capital, just after it had raised $221 million in its Series B round.
The little cloud provider is reportedly valued at $2 billion, and recently [18]revealed it had managed to get one of its customers up and running on Nvidia's latest H100 chips. ®
Get our [19]Tech Resources
[1] https://www.theregister.com/2023/06/02/ai_drone_simulation/
[2] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=2&c=2ZH4Gnu2ZntcMMpD@afJRVAAAAEQ&t=ct%3Dns%26unitnum%3D2%26raptor%3Dcondor%26pos%3Dtop%26test%3D0
[3] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44ZH4Gnu2ZntcMMpD@afJRVAAAAEQ&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[4] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33ZH4Gnu2ZntcMMpD@afJRVAAAAEQ&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[5] https://newsroom.gettyimages.com/en/getty-images/getty-images-statement
[6] https://www.reuters.com/technology/getty-asks-london-court-stop-uk-sales-stability-ai-system-2023-06-01/
[7] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44ZH4Gnu2ZntcMMpD@afJRVAAAAEQ&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[8] https://www.challengergray.com/blog/may-2023-layoffs-jump-on-tech-retail-auto-ytd-hiring-lowest-since-2016/
[9] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33ZH4Gnu2ZntcMMpD@afJRVAAAAEQ&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[10] https://www.theregister.com/2023/05/18/bt_redundancies_2030/
[11] https://www.bloomberg.com/news/features/2023-05-31/jpmorgan-s-push-into-finance-ai-has-wall-street-rushing-to-catch-up
[12] https://www.theregister.com/2023/06/01/foxconn_ai_server/
[13] https://www.theregister.com/2023/06/02/microsoft_cortana_ads/
[14] https://www.theregister.com/2023/06/02/amazon_is_using_ai_software/
[15] https://www.theregister.com/2023/06/02/amazon_is_using_ai_software/
[16] https://www.cnbc.com/2023/06/01/microsoft-inks-deal-with-coreweave-to-meet-openai-cloud-demand.html
[17] https://www.businesswire.com/news/home/20230531005132/en/CoreWeave-Secures-200-Million-Series-B-Extension-Bringing-Total-Round-to-421-Million
[18] https://www.coreweave.com/blog
[19] https://whitepapers.theregister.com/
"misspoke" is now Miss Poked, poor girl
It's worth reviewing what we see these days compared to the environment 50 to 60 years ago in the UK when virtually every problem we see these days did not exist, as a kid after WW2 I saw people with guns (I had a pistol too) but I never heard of anyone getting shot or stabbed ... I still have my knife that I wore every week to my Boy Scout meeting, everyone there wore a knife and nobody ever pulled it out to threaten anyone or worried that we were all throwing our knives at the dartboard. Back then, after the war, people were only concerned with helping other people ... what we call these days as "social-media" didn't exist, AI didn't exist, all we did back then was listen to the BBC radio.
So there have been a lot of changes in the last 60 years ... it's just our guess about the cause of today's problems. Is posting our guesses on the social-media helping make things worse?
Which Part?
Which letter of the word "Simulation" do people not understand?
Have IQ's really dropped this low?
Re: Which Part?
It's worse than that. There was not even a simulation. It was basically one guy saying "Hey, what if a drone figured out its operator can abort its mission, and decided to take him out to prevent that?", and somehow, from just that, we got headlines saying that someone was killed by a rogue AI.
This is about on the same level as people running out screaming from the cinema where they showed a train running towards the camera, except that some of the headlines the next day are suggesting that someone actually got run over by that train.
Journalism today is a sad, sad state of affairs. At least The Register never suggested that reality was involved, and amended the article as soon as the truth was made clear.
Re: Which Part?
To me, "simulation" means there was an actual human operating the controls of a software robot shooting at virtual things, to model what might go wrong if there were an actual physical robot shooting at real-world targets. And "thought experiment" means there was no human operator and no software robot, just somebody asking what might have gone wrong if they had bothered to do a simulation.
Edit: @Filippo was there first.
Not surprised
I spoke to a military chap who told me he was in control of a drone in Afghanistan (I think that’s where it was). It locked on to a guy herding goats. The military chap lost all control of the drone and could do nothing to stop it killing the chap and his goats.
That never made the news. Can’t imagine why
Sure Jan
Ann Stefanek, a spokesperson from the US Air Force, however, told The Register: "The Department of the Air Force has not conducted any such AI-drone simulations and remains committed to ethical and responsible use of AI technology. This was a hypothetical thought experiment, not a simulation."
Sure Jan. What with the Royal Aeronautical Society having a record of exaggeration and sensationalism and all.
https://en.wikipedia.org/wiki/Royal_Aeronautical_Society
"Challenger, Gray & Christmas reckon nearly 4,000 jobs have been replaced with the automated technology."
Maybe it's a sensitive topic for them. Writing reports like this must be a prime target for AI.