Jump To: 
What Exactly is AI?
Are We Using AI?
Does AI Work?
Is AI Ready for EDiscovery?
Is AI Safe?
Conclusion

AI was the big buzzword at Legal Tech last month. As Craig Ball said about what he called the “New New Thing”, “This was the year Generative AI ate Legal Week” and Altorney Co-Founder Rachi Messing referred to ChatGPT as the “tech du jour”.  According to Dean Brown, CEO of conference sponsor IPRO, “Generative AI was on every attendee’s mind, and while ChatGPT is not ready to support eDiscovery, it has helped to bring AI to the forefront of the conversation in our industry.” (https://ediscoverytoday.com/2023/03/27/legalweek-2023-observations-from-attendees-ediscovery-trends/ )

Remember that second clause, we’ll come back to it later. Because I had originally intended to write a recap of the AI land rush several days after the show, but the stampede of articles, interviews, press releases, blog posts and webinars has continued unabated. I’m writing this on the morning of Monday, April 10th so before any more AI stories hit my inbox, let’s start that conversation.

1. WHAT EXACTLY IS AI?

When people perform some tasks, they typically combine a variety of cognitive processes, including logical thinking, strategy development, planning, and decision-making. This process is what we commonly call human intelligence. Good examples are playing chess, writing computer codes, translating languages and even driving a car.

When scientists or engineers automate such an activity, it is often called AI. Artificial general intelligence (AGI) is sometimes also called strong AI, full AI, or general intelligent action, and is the attempt of an intelligent agent to understand or learn any intellectual task that human beings perform. Interestingly, an intelligent agent to a scientist is anything that perceives its environment, takes actions autonomously in order to achieve goals, and can improve its performance by learning or somehow acquiring knowledge.

Of course, by that definition, both thermostats and people are considered an example of an intelligent agent. So we have the first paradox of our discussion: the definition of artificial intelligence includes a sub-definition that includes humans. So how can it be artificial?

Some academic sources reserve the term "strong AI" for computer programs that experience sentience or consciousness and contrast it with weak AI (or narrow AI), which is not intended to have general cognitive abilities but is designed to solve exactly one problem. Some academic sources even use "weak AI" to refer more broadly to any programs that do not experience consciousness or do not have a mind in the same sense people do.

AI projects have been underway for years.  In fact, René Descartes foreshadowed some aspects of it in his 1637 Discourse on the Method of Rightly Conducting One's Reason and of Seeking Truth in the Sciences. And researchers in the United Kingdom kept exploring "machine intelligence" until the formal Dartmouth conference of 1956 which is widely considered the "birth of AI".

One of the early attempts to test AI was introduced by Alan Turing in his 1950 paper "Computing Machinery and Intelligence" while working at the University of Manchester. The Turing test proposed that a human evaluator would judge natural language conversations between a human and a machine designed to generate human-like responses. The evaluator would be aware that one of the two partners in conversation was a machine, and if the evaluator could not reliably tell the machine from the human, the machine would be said to have passed the test.

The test results would not depend on the machine's ability to give correct answers to questions, only on how closely its answers resembled those a human would give

More recently, lawyers entered the discussion in a manner that I called “Fight Club for Lawyers.” (https://www.digitalwarroom.com/blog/is-ai-the-fight-club-of-legal-technology-and-ediscovery) and to prove that lawyers don’t have the high ground on how to complicate a discussion, the National Institute for Standards (NIST) attempted to develop an AI standard in 2021 wherein they coined an acronym for AI user trust characteristics in their model called Perceived System Trustworthiness. That’s right. PST. A phrase we already use in legal data management.  I called this effort  “looking at the wrong end of the horse.” ( https://technogumbo.blog/2021/07/25/is-the-new-nist-standard-for-ai-looking-at-the-wrong-end-of-the-horse/ ).

And then to further complicate matters, some speakers at LegalTech often referred to “generative AI”.  This form of AI can create original articles, essays, images, music, and yes, code, by building on patterns it finds in existing text, audio files, images, and software.

Generative AI  gained much attention for its ability to produce images and text, to the point that late last summer, an  AI-generated image won an art contest.  Then in November 2022, came the debut of ChatGPT, a next-generation chatbot (remember that phrase, it’s going to be important in a moment) created by Open AI, a research lab in San Francisco.

The problem with all this development is … well, I’ll discuss the problems in more detail below. First, let’s continue with the assessment of what AI is and how we are using it.

2. ARE WE USING AI?

The pace of change in generative AI right now is stunning.  OpenAI released ChatGPT to the public just four months ago. It took only five days to generate a million users and two months to reach 100 million users. (TikTok, the internet’s previous instant sensation, took nine.) Google, scrambling to keep up, rolled out Bard, its own AI chatbot, and there are already various ChatGPT clones as well as new plug-ins to make the bot work with popular websites like Expedia and OpenTable.

Then in early 2023, Microsoft announced a $10B investment in OpenAI, the creator of the ChatGPT chatbot. It has already integrated the technology into its Bing search engine. Although the new Bing is available for everyone, it is still in the preview phase, and users must sign up to test it.

The first Microsoft application to integrate a natural language tool predates the excitement over the public release of ChatGPT. GitHub Copilot is an online repository for computer code, owned by Microsoft and used by over 100 million software developers. It has been available since 2021 and uses OpenAI's Codex engine – a modified version of GPT-3 specifically trained to write code – to provide autocomplete features for coders.

In effect, this will mean that it can “bring to life” some of its most well-known and widely used applications and tools. Once integrated with natural language AI, we can (in theory) converse with the application in a very similar way as we could do if it were an actual person.

Can you say Star Trek? Sure you can. Seem far-fetched? Well, CEO Satya Nadella has said that “Every product of Microsoft will have some of the same AI capabilities to completely transform the product.” (https://www.theverge.com/2023/1/17/23558530/microsoft-azure-openai-chatgpt-service-launch )

GPT-4, the new version of the OpenAI model released last month, is both more accurate and “multimodal,” handling text, images, video, and audio all at once. Image generation is advancing at a similarly frenetic pace with a frenzied attention to the programs’ ability to generate highly accurate deep fake images which are difficult to verify as fake.

But according to a Goldman Sachs report, we’re using it so much we’re going to go all Shakespeare and replace the lawyers. Or at least a lot of them. https://www.law.com/corpcounsel/2023/03/29/generative-ai-could-automate-almost-half-of-all-legal-tasks-goldman-sachs-estimates/

But AI has been around long before lawyers started using it. Some examples of AI which you have been using for years, especially in your cell phone, include:

• Maps and Navigation. AI has drastically improved traveling. ...
• Facial Detection and Recognition. ...
• Text Editors or Autocorrect. ...
• Search and Recommendation Algorithms. ...
• Digital Assistants. ...
• Social Media. ...
• E-Payments
• Chatbots
• Real estate and property management

More specifically in our profession, we have been using AI for

• Due Diligence. Litigators conduct due diligence using AI tools to uncover background information. ...
• Predictive technology. ...
• Legal Analytics. ...
• Document Automation. ...
• Intellectual Property (IP). ...
• Electronic Billing
•  Access to justice self help
• Chatbots

Going further, Ralph Losey recently discussed a number of legal products with Generative Large Language Models (LLMs) used as their basis including:

• Legal research
• Draft docs
• Analysis, incl. arguments
• Due diligence
• Contract review
• #eDiscovery re-imagined
• Case management
• ADR
• Legal ed + training
• Client communications

https://e-discoveryteam.com/2023/03/25/ten-ways-llm-models-such-as-chatgpt-can-be-used-to-assist-lawyers/

Several law firm examples of these uses include what DLA Piper's new chief data scientist Bennett Borden calls the “Iron Man project.”, a phrase I love because it connotes the enhancement of human skills that Bennet describes when he says “We aren’t trying to replace lawyers.  We’re trying to extend and augment their natural and trained legal acumen by surrounding them with impressive tech.”

https://www.law.com/legaltechnews/2023/03/21/dla-pipers-chief-data-scientist-firms-resisting-ai-are-dinosaurs-before-the-meteorite-hit/?kw=DLA%20Piper%27s%20Chief%20Data%20Scientist:%20Firms%20Resisting%20AI%20Are%20%27Dinosaurs%20Before%20the%20Meteorite%20Hit%27&utm_source=email&utm_medium=enl&utm_campaign=dailyalert&utm_content=20230322&utm_term=ltn&oly_enc_id=8797H6357367G2S

SIDENOTE Yes, I wish ALM would start using AI to deploy manageable links in their articles. Their annoying practice of these 5-to-6-line links has been going on for years and is easily replaced.  Hello. McFly????

Another early adopter of generative AI was Jay Edelson, the founder and CEO of Chicago-based plaintiffs firm Edelson, PC.  Recently, the firm began using ChatGPT to draft initial press releases as well as to power an internal chief happiness officer named “Chatty.”

In specific legal applications, DISCO has announced Cecilia, a chatbot designed to allow lawyers to ask questions and receive answers with specific citations to supporting evidence in large-scale, private DISCO Ediscovery databases. Unlike chatbots that answer based on only public data sources or the Internet, DISCO Cecilia is designed to cite evidence drawn from the user’s private documents in DISCO Ediscovery.

Relativity announced a patent AI system at LegalWeek 2022 and in this year’s show, CaseText Co-Counsel revealed OpenAI to power tasks such as legal research memo drafting, deposition preparation and document review. Haystack ID touted next-generation AI in ReviewRight Match AI, an advanced document reviewer ranking and selection technology and ReviewRight Staff, a global document reviewer marketplace that matches certified candidates’ qualifications and performance with specific staffing needs.

Also at LegalWeek 2023 were my old friends Rachi and Shimmy Messing, not exhibiting but meeting with people to discuss their new product Altorney. This is a SaaS platform that uses AI to bring a gig marketplace into the legal space with a single source for the recruitment and management of document review attorneys and other qualified reviewers.

And just before LegalWeek 2023, OpenAI announced that Chat GPT 4 passed a bar exam. Should we be surprised?

Well, no.  Remember that computers have been beating people at chess for years. Chess. A seemingly infinite number of moves which depend on the situation to perform correctly. But are they really infinite?

The Shannon Number represents all the possible move variations in the game of chess. It is estimated there are between 10¹¹¹ and 10¹²³ positions (including illegal moves) in chess. If you rule out illegal moves, that number drops significantly to 10^40.  But still far more than a single person can calculate in a reasonable amount of time.

An even more complex game is Go. A company called Deepmind released a product called AlphaGo Zero which, according to the company, learned the game by playing thousands of matches with amateur and professional players. It then began playing against itself and then by playing against the strongest player in the world, the previous version from Deepmind.

The computer program accumulated thousands of years of human knowledge during a period of just a few days. It also developed unconventional strategies and creative new moves, going on to beat several human World Go Champions.

The point? It shows that AI can enhance human ingenuity rapidly and have ‘creative moments’. When dealing with vast amounts of information, especially data involving mathematics, humans can become overwhelmed and quickly become tired. AI doesn’t have those problems.

Lawyers already know from the extensive research into TAR and predictive coding tools used in eDiscovery. But even though most lawyers and law students are aware of generative AI tools such as ChatGPT and their potential advantages, a survey released at LegalWeek by LexisNexis shows few are currently using it for their work.

Only 14% of lawyers and 10% of law students said they have not heard of generative AI tools such as ChatGPT. Still only a minority of either group—36% of lawyers and 44% of law students—said they have used such tools either personally or professionally.

What’s more, 81% of legal respondents noted they are not using such tools currently, while 9% said they use it weekly or monthly, and 2% said they are using such technology daily.

“Generative AI and large language models have tremendous potential to transform the way legal work is done,” said Mike Walsh, CEO of LexisNexis Legal & Professional, in a press release announcing the survey.

Whether such impact is welcome by the market, however, is another story. Only 14% of respondents said generative AI tools will have a positive impact on the practice of law, while the same amount believed it would have a negative impact. Almost two-thirds, 63%, said the impact would be mixed.

Additionally, 54% of respondents had some concerns or questions about the ethical implications of generative AI on the practice of law, while 28% had significant concerns.

To be sure, generative AI isn’t just expected to impact the practice of law. A majority of lawyers, 52%, and legal students, 61%, also predicted generative AI will change legal education as well.

An earlier study in the fall of 2022, however, The Artificial Intelligence and Machine Learning Report from the Berkeley Research Group (BRG), Relativity and ACEDS, found a distinct difference between law firms and legal departments. It also showed that the biggest hurdles to embracing both AI and machine learning are cost and meaningful tech education.

Does the lack of education lead to market confusion?  Well to be polite about it, obfuscation would be a better description.

In their 2021 publication, The eDiscovery Medicine Show, (Ohio State Technology Law Journal 18:1 (2021)), well known search technology experts Maura Grossman and her husband, Gordon Cormack, wrote:
 

“eDiscovery methods, like therapeutics, are amenable to scientific evaluation. But practitioners and their “experts,” vendors, and clients often ignore empirical evidence, citing instead existing or past practice to justify, for example, culling electronically stored information (“ESI”) using untested search terms, establishing neither their necessity nor their efficacy. Or, they use pseudo-science to promote various potions marketed as “Artificial Intelligence,” “AI,” “technology-assisted review,” or “TAR.” Or, they employ pseudoscience and various logical fallacies to impugn scientific studies that contradict their claims. Or, they point to the often-cited Sedona Principle 68 as justification to do whatever they please. Or, sometimes, even all of the above.”

“Trade shows and other “educational” activities sponsored by vendors promote their wares, complete with pseudo-scientific results, testimonials, sponsored receptions, prizes, and hospitality suites. The Continuing Legal Education (“CLE”) industry and the trade press often echo these testimonials, failing to discriminate between practice and sound practice—let alone best practice—or between science and pseudo-science. So far, neither the courts nor any other authority has taken up the mantle, leaving parties to fend for themselves in the eDiscovery Wild West.”

So given what Maura & Gordon say, the real question becomes “Does AI work?”

3. DOES AI WORK?

Clearly for many tasks, as the lists above showed us, the answer is yes. In an interview with Above The Law, James Michalowicz, senior manager of legal-ops business performance at TE Connectivity, said he felt that findings of the Goldman Sachs report are generally accurate when it comes to the replacement of legal tasks at hands of AI.

“My first reaction [to the report] was ‘Oh boy, here we go again,’” Michalowicz said. “It’s kind of like, there’s a disruption and the next thing that happens, especially in the legal industry, is pushback and resistance and questioning of [the] data. This is like my fifth or sixth iteration of this [which] I have experienced.”

In the past, this sort of reaction and fear has happened with contract attorneys, then predictive coding and technology assisted review (TAR), then outsourcing of legal tasks overseas and now with AI, Michalowicz noted.

For him, AI and even the most advanced generative AI is more likely to ease the burden of overloaded legal staffs and increase productivity as it “allows you to free up capacity to handle some of the more strategic work”.

But not everyone agrees.

The signatories of the Future of Life Institute's recent open letter — which include SpaceX CEO Elon Musk, Apple co-founder Steve Wozniak, and onetime presidential candidate Andrew Yang — called for a six-month pause on AI advancement to take stock.

Why? Well, the web demo of Alpaca, a small AI language model based on Meta's LLaMA system, has been taken down offline by researchers at Stanford University due to safety and cost concerns. Like all other language models, Alpaca is prone to generating misinformation, a property often described as hallucination. Offensive text is another common output.

"Hallucination in particular seems to be a common failure mode for Alpaca, even compared to text-davinci-003," the researchers noted. In some examples, the model failed to recall the capital of Tanzania correctly and produced false technical information. 

Like all other language models, Alpaca is prone to generating misinformation, a property often described as hallucination. Offensive text is another common output.

"Hallucination in particular seems to be a common failure mode for Alpaca, even compared to text-davinci-003," the researchers noted. In some examples, the model failed to recall the capital of Tanzania correctly and produced false technical information. https://www.theregister.com/2023/03/21/stanford_ai_alpaca_taken_offline/

Here’s a good example. A recent ChatGPT exercise asked the program “give me a picture of a gamer on a skateboard.”  The result?

And in a recent post on-line, Steve Goldstein, the Global Director of Practice Support at Squire Patton Boggs and self-proclaimed “AI Believer” recounted how a ChatGPT query about himself revealed that he graduated from Harvard Law School and was a partner at SPG. Neither is true.

So, this naturally begs the following question …

4. IS AI READY FOR EDISCOVERY?

Remember the quote from IPRO CEO Dean Brown at the top of this post? Dean said point blank, “ChatGPT is not ready to support eDiscovery.”

And in a separate post, Kevin Clark, Discovery Counsel & Vice President of Analytics and Review was asked:

Q: Will ChatGPT-4 be a useful analytical tool in document review for litigation or investigations when performing eDiscovery?

A: ChatGPT-4 could potentially be a useful analytical tool in document review for litigation or investigations when performing eDiscovery. As an AI language model, it has the potential to assist with tasks such as document classification, entity recognition, and text summarization, which are all important in eDiscovery. However, the effectiveness of ChatGPT-4 would depend on a number of factors, including the quality and quantity of the data it is trained on, the specific tasks it is designed to perform, and the expertise of the individuals who train and use it.’
…

“However, it is important to note that ChatGPT-4 (or any AI model) should not be relied upon as the sole method of document review in eDiscovery. Human review and oversight is still necessary to ensure that the AI-generated results are accurate and relevant to the specific case or investigation. Additionally, there may be limitations to what an AI model like ChatGPT-4 can identify, particularly in cases where contextual understanding and human judgment are required.”

Research scientists are also frustrated because there is extreme secrecy around Open AI's ChatGPT 4. “All of these closed-source models, they are essentially dead-ends in science," Nature quoted Sasha Luccioni, a research scientist specializing in climate at HuggingFace, an open-source-AI community. “They [OpenAI] can keep building upon their research, but for the community at large, it’s a dead end."

Without access to the data used for training, OpenAI’s assurances about safety fall short for Luccioni. “You don’t know what the data is. So you can’t improve it. I mean, it’s just completely impossible to do science with a model like this," she says.

The mystery about how GPT-4 was trained is also a concern for van Dis’s colleague at Amsterdam, psychologist Claudi Bockting. “It’s very hard as a human being to be accountable for something that you cannot oversee," she says. “One of the concerns is they could be far more biased than for instance, the bias that human beings have by themselves." Without being able to access the code behind GPT-4 it is impossible to see where the bias might have originated, or to remedy it, Luccioni explained.

( https://www.livemint.com/news/world/its-a-dead-end-researchers-share-their-opinion-on-chatgpt4-11679154165405.html )

Or explain it to a judge. This lack of transparency and ability to explain results may be deadly for AI usage in eDiscovery. And they beg the further question…..

5.  IS AI SAFE?

In the post-LegalTech article note above, Dave Ruel, VP of Product, Hanzo, was quoted as saying:

“AI is being rapidly adopted into many facets within E-Discovery, but the caution signs are starting to also appear with questions around how AI is not only used but how it is reported on and audited across the EDRM model.”

Elon Musk, who co-founded OpenAI in 2015 and eventually left its board in 2018, tweeted as far back as 2002 that:

“I have no control & only very limited insight into OpenAI,” adding his confidence in its safety was “not high.” He then paused OpenAI’s access to the Twitter database for training, saying he needed to know more about the A.I. creator’s “governance structure & revenue plans going forward. OpenAI was started as open-source & non-profit. Neither are still true.”

https://fortune.com/2023/03/28/elon-musk-tesla-twitter-ceo-feud-bill-gates-microsoft-over-ai-chatgpt-openai-talks/

Even Bill Gates, who has invested in OpenAI and did not sign the FLI letter, expressed concerns during a Q&A session on Reddit in January 2015, saying  “I am in the camp that is concerned about super intelligence. First, the machines will do a lot of jobs for us and not be super intelligent. That should be positive if we manage it well. A few decades after that though the intelligence is strong enough to be a concern. I agree with Elon Musk and some others on this and don’t understand why some people are not concerned.”

And a group of well-known AI ethicists who did not sign the six month pause letter did so because they say it focused on hypothetical future threats when real harms are attributable to misuse of the technology today.

Timnit Gebru, Emily M. Bender, Angelina McMillan-Major and Margaret Mitchell are all major figures in the domains of AI and ethics, known for both their work and  for being pushed out of Google over a paper criticizing the capabilities of AI. They are currently working together at the DAIR Institute, a group studying AI-associated harms.

They published a rebuke to the first letter, saying:

“The choice to worry about a Terminator- or Matrix-esque robot apocalypse is a red herring when we have, in the same moment, reports of companies like Clearview AI being used by the police to essentially frame an innocent man. No need for a T-1000 when you’ve got Ring cams on every front door accessible via online rubber-stamp warrant factories.”

Strong words but in fact an algorithm sent a black man to jail in Atlanta over a warrant from Louisiana, a state he'd never in fact visited. A facial recognition tool identified Randall Reid as a suspect in the theft of luxury purses in Baton Rouge and it took several weeks of incarceration before the police admitted they were mistaken. ( https://gizmodo.com/facial-recognition-randall-reid-black-man-error-jail-1849944231 )

Facial recognition is, in fact, far from perfect. Numerous studies show the technology is especially inaccurate when identifying people of color and women compared to identifications of white men and is a troubling trend, according to Clare Garvie, training resource counsel with the National Association of Criminal Defense Lawyers.

In fact, a 2021 study by the National Institute of Standards and Technology found all the top algorithms can identify airport passengers with photos in the system more than 99% of the time on the first appearance before a camera. But that study and others also found continuing bias in differing rates of false positives and negatives by race and gender.

Joy Buolamwini, a facial recognition researcher at the Massachusetts Institute of Technology, has uncovered racial and gender bias in facial analysis tools sold by companies such as Amazon. Some critics have argued that the technology is so perilous in government hands that it should be banned.

The Electronic Privacy Information Center argued this year that facial recognition is “inherently dangerous,” enabling “comprehensive public surveillance.” “It’s a powerful surveillance tool that can easily be expanded without people’s knowledge,” said Jeramie Scott, senior counsel with EPIC. (https://democrats-homeland.house.gov/activities/hearings/assessing-cbps-use-of-facial-recognition-technology )

Sound alarmist? Well not to the European community.

Last week, Garante, the Italian Data Protection watchdog, ordered OpenAI to temporarily cease processing Italian users’ data amid a probe into a suspected breach of Europe’s strict privacy regulations. The regulator cited a data breach at OpenAI which allowed users to view the titles of conversations other users were having with the chatbot. Garante also flagged worries over a lack of age restrictions on ChatGPT, and how the chatbot can serve factually incorrect information in its responses.

In a public statement, they said that there “appears to be no legal basis underpinning the massive collection and processing of personal data in order to ‘train’ the algorithms on which the platform relies.”

CONCLUSION

So, the problem with AI is not just defining what you mean when you say AI but also defining what you are doing with it and then supervising it to eliminate problems with small mistakes, major errors and gaffes so large they are called hallucinations. It’s also a question of bias, invasion of privacy and lack of transparency.  All issues any one of which alone would be problematic but in combination portend even larger problems for AI in the eDiscovery world.

Dave Lewis Co-Founder and Chief Scientific Officer at Redgrave Data recently posted on LinkedIn in response to a story about a web site with a chatbot providing various means of suicide to people on line:

“A tragic and sadly unsurprising consequence of AI hype. IT IS JUST A STATISTICAL MODEL OF WORDS STRUNG TOGETHER. Putting general purpose "conversational" interfaces on these systems and giving them cute names is deeply irresponsible. Let's have a 6-month moratorium not on AI research, but rather on hyping fake AGI dangers over the ever green danger of mispurposed and over promised software.”

https://www.linkedin.com/feed/update/urn:li:activity:7049805851808727040/

Sophie Hackford, a futurist and global technology innovation advisor for American farming equipment maker John Deere, told CNBC’s “Squawk Box Europe” recently that:

“Technology is here to serve us. it’s there to make our cancer diagnosis quicker or make humans not have to do jobs that we don’t want to do.”

“We need to be thinking about it very carefully now, and we need to be acting on that now, from a regulation perspective,” she added.

And Ralph Losey, in the article quoted above, has said:

“The assistance of carefully prompted and quality controlled ChatGPT-4 software can, if used correctly, dramatically improve the legal profession. But use caution and be wary, especially in these early days. Keep your new AI helpers on a short leash. They can make some mistakes, both big and small. As always with new tech, beware of exaggerated claims and vendor fluff.

One recent set of articles from the FTC posted by Rob Robinson seem to indicate that at least one regulatory agency is taking note but I wonder if the recent trend towards privacy regulation in various states will cause the agencies in those states to see what is happening in Europe with OpenAI and privacy issues.

( https://complexdiscovery.com/holding-providers-accountable-considering-ai-in-ediscovery-service-and-software-provider-marketing-and-messaging/?utm_source=ComplexDiscovery+Opt-In+Subscriber+Update&utm_campaign=5aacf311d7-EMAIL_CAMPAIGN_041023_FTC_ON_AI&utm_medium=email&utm_term=0_c76d993979-5aacf311d7-274470677&mc_cid=5aacf311d7&mc_eid=cc2c2be60f

Well known commentator Gregory Bufithis, founder of both The Project Counsel Group and Luminative Media, had a recent article called AI is not the only system that hallucinates  in his online column “Thoughts Over My Afternoon Coffee”.  The article discusses a Wired magazine article about the letter calling for a halt to ChatGPT development and he references this quote from Wired:

“Instead of halting research, we need to improve transparency and accountability while developing guidelines around the deployment of AI systems. Policy, research, and user-led initiatives along these lines have existed for decades in different sectors, and we already have concrete proposals to work with to address the present risks of AI.”

I’ve remarked before that in working with computers in any arena, humans are slow but smart while computers are fast but dumb. Someone commented recently that we should never mistake the fast accumulation of data for intelligence.

As I wrote last year, let’s keep the attorney in AI.  

 https://gulfltc.org/wp-content/uploads/2023/03/Keep-Attorney-In-AI.pdf

About the Author...

Tom O'Connor
Gulf Coast Legal Technology Center
E-Discovery Committee Chair