Tag Archives: #Artifical Intelligence

AI hallucinations and the law

Posted on by
Reply

Several years ago, I blogged about the role of technology within the legal profession. One development I noted was the nascent use of AI to help test the merits of a case before it goes to trial, and to assess the likelihood of winning. Not only might this prevent potentially frivolous matters coming to trial, it would also reduce court time and legal costs.

More recently, there has been some caution (if not out and out scepticism) about the efficacy of using AI in support of legal research and case preparation. This current debate has been triggered by an academic paper from Stanford University that compared leading legal research tools (that claim to have been “enhanced” by AI) and ChatGPT. The results were sobering, with a staggering number of apparent “hallucinations” being generated, even by the specialist legal research tools. AI hallucinations are not unique to legal research tools; nor to the AI tools and the Large Language Model (LLMs) they are trained on, as Stanford has previously reported. While the academic paper is awaiting formal publication, there has been some to-and-fro between the research authors and at least one of the named legal tools. This latter rebuttal rightly points out that any AI tool (especially a legal research and professional practice platform) has to be fit for purpose, and trained on appropriate data.

Aside from the Stanford research, some lawyers have been found to have relied upon AI tools such as ChatGPT and Google Bard to draft their submissions, only to discover that the results have cited non-existent precedents and cases – including in at least one high-profile prosecution. The latest research suggests that not only do AI tools “imagine” fictitious case reports, they can also fail to spot “bad” law (e.g., cases that have been overturned, or laws that have been repealed), offer inappropriate advice, or provide inaccurate or incorrect legal interpretation.

What if AI hallucinations resulted in the generation of invidious content about a living person – which in many circumstances, would be deemed libel or slander? If a series of AI prompts give rise to libelous content, who would be held responsible? Can AI itself be sued for libel? (Of course, under common law, it is impossible to libel the dead, as only a living person can sue for libel.)

I found an interesting discussion of this topic here, which concludes that while AI tools such as ChatGPT may appear to have some degree of autonomy (depending on their programming and training), they certainly don’t have true agency and their output in itself cannot be regarded in the same way as other forms of speech or text when it comes to legal liabilities or protections. The article identified three groups of actors who might be deemed responsible for AI results: AI software developers (companies like OpenAI), content hosts (such as search engines), and publishers (authors, journalists, news networks). It concluded that of the three, publishers, authors and journalists face the most responsibility and accountability for their content, even if they claimed “AI said this was true”.

Interestingly, the above discussion referenced news from early 2023, that a mayor in Australia was planning to sue OpenAI (the owners of ChatGPT) for defamation unless they corrected the record about false claims made about him. Thankfully, OpenAI appear to have heeded of the letter of concern, and the mayor has since dropped his case (or, the false claim was simply over-written by a subsequent version of ChatGPT). However, the original Reuters link, above, which I sourced for this blog makes no mention of the subsequent discontinuation, either as a footnote or update – which just goes to show how complex it is to correct the record, since the reference to his initial claim is still valid (it happened), even though it did not proceed (he chose not to pursue it). Even actual criminal convictions can be deemed “spent” after a given period of time, such that they no longer appear on an individual’s criminal record. Whereas, someone found not guilty of a crime (or in the mayor’s case, falsely labelled with a conviction) cannot guarantee that references to the alleged events will be expunged from the internet, even with the evolution of the “right to be forgotten“.

Perhaps we’ll need to train AI tools to retrospectively correct or delete any false information about us; although conversely, AI is accelerating the proliferation of fake content – benign, humourous or malicious – thus setting the scene for the next blog in this series.

Next week: AI and Deep (and not so deep…) Fakes

 

 

 

 

AI and the Human Factor

Posted on by
Reply

Earlier this month, I went to the Melbourne premiere of “Eno”, a documentary by Gary Hustwit, which is described as the world’s first generative feature film. Each time the film is shown, the choice and sequencing of scenes is different – no two versions are ever the same. Some content may never be screened at all.

I’ll leave readers to explore the director’s rationale for this approach (and the implications for film-making, cinema and streaming). But during a Q&A following the screening, Hustwit was at pains to explain that this is NOT a film generated by AI. He was also guarded and refrained from revealing too much about the proprietary software and hardware system he co-developed to compile and present the film.

However, the director did want to stress that he didn’t simply tell an AI bot to scour the internet, scrape any content by, about or featuring Brian Eno, and then assemble it into a compilation of clips. This documentary is presented according to a series of rules-based algorithms, and is a content-led venture curated by its creator. Yes, he had to review hours and hours of archive footage from which to draw key themes, but he also had to shoot new interview footage of Eno, that would help to frame the context and support the narrative, while avoiding a banal biopic or series of talking heads. The result is a skillful balance between linear story telling, intriguing juxtaposition, traditional interviews, critical analysis, and deep exploration of the subject. The point is, for all its powerful capabilities, AI could not have created this film. It needed to start with human elements: innate curiosity on the part of the director; intelligent and empathetic interaction between film maker and subject; and expert judgement in editing the content – as a well as an element of risk-taking in allowing the algorithm to make the final choices when it comes to each screened version.

That the subject of this documentary is Eno should not be surprising, either. He has a reputation for being a modern polymath, interested in science and technology as well as art. His use of Oblique Strategies in his creative work, his fascination with systems, his development of generative music, and his adoption of technology all point to someone who resists categorisation, and for whom work is play (and vice versa). In fact, imagination and play are the two key activities that define what it is to be human, as Eno explored in an essay for the BBC a few years ago. Again, AI does not yet have the power of imagination (and probably has no sense of play).

Sure, AI can conjure up all sorts of text, images, video, sound, music and other outputs. But in truth, it can only regurgitate what it has been trained on, even when extrapolating from data with which it has been supplied, and the human prompts it is given. This process of creation is more akin to plagiarism – taking source materials created by other people, blending and configuring them into some sort of “new” artefact, and passing the results off as the AI’s own work.

Plagiarism is neither new, nor is it exclusive to AI, of course. In fact, it’s a very natural human response to our environment: we all copy and transform images and sounds around us, as a form of tribute, hommage, mimicry, creative engagement, pastiche, parody, satire, criticism, acknowledgement or denouncement. Leaving aside issues of attribution, permitted use, fair comment, IP rights, (mis)appropriation and deep fakes, some would argue that it is inevitable (and even a duty) for artists and creatives to “steal” ideas from their sources of inspiration. Notably, Robert Shore in his book about “originality”. The music industry is especially adept at all forms of “copying” – sampling, interpolation, remixes, mash-ups, cover versions – something that AI has been capable of for many years. See for example this (limited) app from Google released a few years ago. Whether the results could be regarded as the works of J.S.Bach or the creation of Google’s algorithm trained on Bach’s music would be a question for Bach scholars, musicologists, IP lawyers and software analysts.

Finally, for the last word on AI and the human condition, I refer you to the closing scene from John Carpenter’s cult SciFi film, “Dark Star”, where an “intelligent” bomb outsmarts its human interlocutor. Enjoy!

Next week: AI hallucinations and the law

 

 

Whose side is AI on?

Posted on by
Reply

At the risk of coming across as some sort of Luddite, recent commentary on Artificial Intelligence suggests that it is only natural to have concerns and misgivings about its rapid development and widespread deployment. Of course, at its heart, it’s just another technology at our disposal – but by its very definition, generative AI is not passive, and is likely to impact all areas of our life, whether we invite it in or not.

Over the next few weeks, I will be discussing some non-technical themes relating to AI – creativity and AI, legal implications of AI, and form over substance when it comes to AI itself.

To start with, these are a few of the questions that I have been mulling over:

– Is AI working for us, as a tool that we control and manage?  Or is AI working with us, in a partnership of equals? Or, more likely, is AI working against us, in the sense that it is happening to us, whether we like it or not, let alone whether we are actually aware of it?

– Is AI being wielded by a bunch of tech bros, who feed it with all their own prejudices, unconscious bias and cognitive limitations?

– Who decides what the Large Language Models (LLMs) that power AI are trained on?

– How does AI get permission to create derived content from our own Intellectual Property? Even if our content is on the web, being “publicly available” is not the same as “in the public domain”

– Who is responsible for what AI publishes, and are AI agents accountable for their actions? In the event of false, incorrect, misleading or inappropriate content created by AI, how do we get to clarify the record, or seek a right of reply?

– Why are AI tools adding increased caveats? (“This is not financial advice, this is not to be relied on in a court of law, this is only based on information available as at a certain point in time, this is not a recommendation, etc.”) And is this only going to increase, as in the recent example of changes to Google’s AI-generated search results? (But really, do we need to be told that eating rocks or adding glue to pizza are bad ideas?)

– From my own experience, tools like Chat GPT return “deliberate” factual errors. Why? Is it to keep us on our toes (“Gotcha!”)? Is it to use our responses (or lack thereof) to train the model to be more accurate? Is it to underline the caveat emptor principle (“What, you relied on Otter to write your college essay? What were you thinking?”). Or is it to counter plagiarism (“You could only have got that false information from our AI engine”). If you think the latter is far-fetched, I refer you to the notion of “trap streets” in maps and directories.

– Should AI tools contain better attribution (sources and acknowledgments) in their results? Should they disclose the list of “ingredients” used (like food labelling?) Should they provide verifiable citations for their references? (It’s an idea that is gaining some attention.)

– Finally, the increased use of cloud-based services and crowd-sourced content (not just in AI tools) means that there is the potential for overreach when it comes to end user licensing agreements by ChatGPT, Otter, Adobe Firefly, Gemini, Midjourney etc. Only recently, Adobe had to clarify latest changes to their service agreement, in response to some social media criticism.

Next week: AI and the Human Factor