News: 1695015014

  ARM Give a man a fire and he's warm for a day, but set fire to him and he's warm for the rest of his life (Terry Pratchett, Jingo)

UK judge rates ChatGPT as 'jolly useful' after using it to help write a decision

(2023/09/18)


AI in brief A judge working at the UK's Court of Appeal has admitted he used ChatGPT to help him write a ruling.

Speaking at a Law Society event, Lord Justice Birss said he turned to ChatGPT to generate a paragraph for court documents in a case related to intellectual property law. Birss said he directly copied and pasted the words into the ruling he wrote, adding that tools like ChatGPT had "great potential," the UK's Telegraph [1]reported .

"I think what is of most interest is that you can ask these large language models to summarize information. It is useful and it will be used and I can tell you, I have used it," he revealed.

[2]

Birss's remarks are thought to be the first reported instance of a British judge admitting to using generative AI software in their work.

[3]

[4]

As such it is controversial, as ChatGPT makes many errors. In the US, two [5]lawyers were heavily criticized for using the chatbot to defend a client in court after judges realized the software had generated false information.

"I'm taking full personal responsibility for what I put in my judgment, I am not trying to give the responsibility to somebody else. All it did was a task which I was about to do and which I knew the answer to and could recognize as being acceptable," Birss argued.

Are VCs cooling on AI startups?

AI-centric chip startups hoping to compete with Nvidia are facing an uphill battle to secure funds from venture capitalist firms.

They have collectively managed to raise $881.4 million this year to the end of August, according to PitchBook data, [6]reported by Reuters. That's a decrease of around $908 million – over 80 percent – compared to the amount of money raised in the first three quarters of last year.

[7]

The data shows that VCs are spending less money backing AI chip startups and making fewer deals: just four startups have received funding this year so far compared to 23 in 2022.

Hardware startups are a riskier proposition than their software-centric cousins because hardware can take years to design and build. Established chip makers have many advantages.

"Nvidia's continued dominance has put a really fine point on how hard it is to break into this market," Greg Reichow, a partner at Eclipse Ventures, said. "This has resulted in a pullback in investment into these companies, or at least into many of them."

[8]The future of the cloud sure looks like it'll be paved in even more custom silicon

[9]Netflix offers up to $900,000 for AI product manager while actors strike for protection

[10]X may train its AI models on your social media posts

Coca-Cola adds AI flavor

Soft drink giant Coca-Cola has created a limited edition variety of Coke with a flavor profile generated using AI.

The drink – dubbed Coca-Cola Y3000 – is described as being "futuristic flavored" and is part of the carbonated colossus's "Creations" series of limited edition varieties. It comes in a silver can with pink, blue, and purple bubbles designed using text-to-image tools. At the bottom it states that it was "co-created with artificial intelligence."

[11]

A spokesperson for the fizzy titan [12]told CNN that the flavor profile was created with help from machine learning. Coca-Cola collected data to see what tastes people associated with the future and turned to software to create different flavor pairings.

So what does it taste like? Normal Coke, apparently – with a twist.

Oana Vlad, senior director of global brand at Coca-Cola, previously told CNN that the drink vendor never discloses what is inside its recipes. "We're never really going to answer that question" he said, at least not in a "straightforward" way. For limited edition varieties Coke's "flavor profile is always, we say, 85 to 90 [percent] Coke. And then that 10 to 15 [percent] twist of something unexpected," he added.

Y3000 will be available in stores starting this week for consumers in the US and Canada.

Uh oh, even The New York Times is getting into AI journalism

The New York Times is hiring a senior editor to bring generative AI tools into its newsroom.

"This editor will be responsible for ensuring that The Times is a leader in GenAI innovation and its applications for journalism," states the [13]ad for the position. "They will lead our efforts to use GenAI tools in reader-facing ways as well as internally in the newsroom."

Putting generative AI to work in newsrooms has proven controversial. Early adopters like Red Ventures' CNET or G/O Media's [14]Gizmodo produced [15]errors – even after human oversight

Such mistakes can be difficult to spot, since AI generates text that is grammatically correct and often convincing. Without deep expertise on a subject, it can be difficult for editors to detect mistakes.

The ad states that whoever gets the gig "will also help shape further guidelines for how GenAI is used by journalists throughout the newsroom, in partnership with the Standards department, taking into account the evolving nature of the technology and its risks."

But The NYT appears to see a role for AI in its newsroom and famously pedantic fact-checking process.

"The editor's primary focus will be on producing a steady stream of projects demonstrating high potential and responsible ways to incorporate GenAI tools into Times journalism and workflows," the ad reads.

The move to incorporate generative AI technology into a top newsroom will no doubt spur others – who don't want to fall behind – to follow suit. ®

Get our [16]Tech Resources



[1] https://www.telegraph.co.uk/business/2023/09/14/british-judge-uses-jolly-useful-chatgpt-to-write-ruling/

[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=2ZQgfxElE8pEjAHplsiDtqwAAA4c&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=44ZQgfxElE8pEjAHplsiDtqwAAA4c&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=33ZQgfxElE8pEjAHplsiDtqwAAA4c&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0

[5] https://www.theregister.com/2023/06/22/lawyers_fake_cases/

[6] https://www.reuters.com/technology/nvidias-dominance-ai-chips-deters-funding-startups-2023-09-11/

[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=44ZQgfxElE8pEjAHplsiDtqwAAA4c&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0

[8] https://www.theregister.com/2023/09/12/the_future_of_the_cloud/

[9] https://www.theregister.com/2023/07/26/netflix_ai_manager/

[10] https://www.theregister.com/2023/09/04/in_brief_ai/

[11] 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=33ZQgfxElE8pEjAHplsiDtqwAAA4c&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0

[12] https://www.cnn.com/2023/09/12/business/coca-cola-y3000/index.html

[13] https://nytimes.wd5.myworkdayjobs.com/en-US/NYT/job/New-York-NY/Newsroom-Generative-AI-Lead_REQ-014906-1

[14] https://www.washingtonpost.com/technology/2023/07/08/gizmodo-ai-errors-star-wars/

[15] https://www.theregister.com/2023/01/19/cnet_reviewing_ai_authored_stories/

[16] https://whitepapers.theregister.com/



Beak praises Stochastic Parrot.

Anonymous Coward

and the Law becomes a Lottery.

TheMaskedMan

"All it did was a task which I was about to do and which I knew the answer to and could recognize as being acceptable,"

Good grief, a sensible judge - I didn't think we had any. Most unusually for a judge, this guy understands that chatGPT et al are tools, to be used but not trusted blindly. If he knows the law anyway, as he obviously does, and is able to see whether it's making shit up or not, why shouldn't he use it to speed up his work? Well done that judge.

Pascal Monett

What reassures me is that it is a judge. He did not simply ask ChatGPT to write his ruling, he consulted it for a specific part, controlled the result and was satisfied enough to include it in his ruling.

That is called using the tool properly, not like those two lawyers who took ChatGPT results and included them without controlling the content.

There's several reasons why lawyers are lawyers and not judges, and this is one.

Fruit and Nutcase

There's several reasons why lawyers are lawyers and not judges, and this is one.

Begs the question why a top flight lawyer like our esteemed home secretary didn't want to become a judge

low_resolution_foxxes

Has anyone got any opinion on what actually happened with the ChatGPT lawyer case? I am unclear how it managed to completely fabricate case law, unless of course the AI was fed with garbage data (intentionally or accidentally).

I am a cynical geek at the best of times. It strikes me that lawyers face perhaps one of the biggest impacts from the AI revolution.

Law and compliance is a staggeringly complicated area and these groups are charging £1000s per hour for their advice. If you can get 95% of the information from chatGPT they stand to get annihilated. So funnily enough it's a bunch of lawyers who first demonstrate the AI is capable of providing false information.

I mean seriously....did he really think you could just file any old nonsense and risk his entire career, on something created by a draft AI system?

I just ran two moderately complicated compliance/legal questions through ChatGPT to gain a feeling for how reliable it can be. It answered one question perfectly, but it had an understandable issue with my 2nd question. It technically provided an incorrect answer - apparently due to the regulation changing in 2022 - but to be fair to ChatGPT it actually opens the conversation by pointing out it was basing its answers on the dataset fed to it in September 2021. So it made me aware of the situation and it is my responsibility to check for directive changes since then.

AI takes the pith

Grunchy

I guess it would be unethical to get an artificially intelligent parser to write pithy remarks for me.

The correct way to use these tools...

jmch

For me, this is the key quote: "All it did was a task which I was about to do and which I knew the answer to and could recognize as being acceptable".

That's no different to asking a junior or an intern to draft you a response. The key is that the judge (a) reviewed the output (b) judged it to be correct and (c) took responsibility for the outcome*. Missing out on any of those steps is what leads to trouble.

*Of course if it's his signature at the bottom, he's responsible for it anyway

Knowing ignorance is strength.
Ignoring knowledge is sickness.

If one is sick of sickness, then one is not sick.
The sage is not sick because he is sick of sickness.
Therefore he is not sick.