ChatGPT is Getting “Lazy” — Let KCLL Do Your Research!
Does AI get SAD?
In the Pacific Northwest we are no strangers to the mood-altering effects of The Big Dark. While some of us embrace the gray skies and rain, others hunker down with our books, coffee, and happy lights to chase the SAD (seasonal affective disorder) away. In what would seem to be more suited to a movie plot or an Onion article, in November people began to comment on tech & social media platforms that ChatGPT4 was getting lazy. Come again? Users reported that ChatGPT4 was refusing to do work and giving much less finished/ detailed responses than previously. Conspiracy theories abound. Is it possible that AI has become so self-aware that it too, falls prey to seasonal affective disorder? The timing is suspicious!
AI Can Definitely Do Sass!
OpenAI, the creator of ChatGPT4 tweeted in response to the online kerfuffle that they had received feedback that the chatbot was reportedly refusing to perform requests or just doing a portion of the work requested and then telling users to finish it themselves. “If the person asks for a piece of code, for instance, it might just give a little information and then instruct users to fill in the rest. Some complained that it did so in a particularly sassy way, telling people that they are perfectly able to do the work themselves, for instance.”1 OpenAI responded that it is working on making GPT-4 less “lazy.”2 If ChatGPT4 is not falling prey to SAD, then it certainly sounds like it’s been talking to some teenagers I know.
What’s causing the sass and the malingering? As Jonathan Franklin, pointed out in last month’s column, a little-discussed fact about AI chat systems is that they are incredibly expensive to run. “It is easy to look at generative AI as ‘free’ because you can go to a website and use it without paying. For now, the costs are being carried by the companies trying to build market share and figure out how to build subscription services that cover their costs. The costs of running the AI hardware are substantial, from the cost of the NVIDIA computer chips, to the cost of water for cooling the servers, to the cost of the employees who are building the tools. If each legal memo generated costs $100 or more, users might not play around with it the way they currently are, so enjoy this introductory period while you can.”
Some are speculating that the bots are being trained to give less detailed answers in an effort to use less computer processing power as a cost saving measure. OpenAI responded on Twitter: “We haven’t updated the model since Nov 11th, and this certainly isn’t intentional. Model behavior can be unpredictable, and we’re looking into fixing it.”
Robot Lawyers Quiet Quitting?
For years there’s been much hand wringing about the robot lawyers coming to take over the legal profession. With the introduction of ChatGPT last November, however, that possibility seemed less theoretical.3 But after having been on the job for a while, is it possible that the robot lawyers are already “quiet quitting?”
From “hallucinated cases” to wrong or incomplete answers, several lawyers have found out the hard way the perils of using AI for legal work without sufficient care. Yes, AI can be an amazing time-saving tool, but only when used with the trust-but-verify approach. This is not a new concept for lawyers, we cite-check cases and update statutes before relying on them.
So, the next time you have a legal research time crunch, don’t abide the sassy, “quiet quitting” robot lawyer, contact the law library. You’ll get the best of both worlds. AI-informed research with Westlaw’s new Precision database coupled with the hands-on expertise of the law library’s research team.
Let KCLL Do Your Research
For those in the know, the law library’s Let Us Do Your Research service is one of the best deals in town. For solo and small firm attorneys, it’s like having an on-demand, low-cost research assistant who comes armed with tens of thousands of dollars of legal research resources, including AI-enabled Precision from Westlaw.
Subscribers submit research projects to us through our website. The first fifteen minutes of research is free as our staff reviews the request and determines whether: (1) the question(s) fall within the scope of our expertise and our information resources; and (2) we can meet the request deadline. Once we determine the question meets these parameters, the subscriber decides whether it makes sense to use the service and how much time she would like us to spend. The current rate for approved Let Us Do Your Research projects is $100 per hour, charged in 15-minute increments. Our researchers will limit the amount of time spent on the question to the predetermined amount. If the question takes less time than projected, we only charge for the time spent on the project. Subscribers receive a detailed research memo that includes references to the resources used, summaries of relevant caselaw as well as the full text of cases relied on in the research memo. This service is available only to subscribers.
Here’s what some of our Let Us Do Your Research service users have said:
I presented a fairly complicated question and didn’t want to spend my time down the rabbit hole. I wanted a pro to help me avoid the hours I would have taken to get to first base. I got a call a day after I sent the question from a gentleman who asked perfect questions and engaged in a fruitful and interesting back and forth on the issue. Three days later I got a clearly written road map memo that got me where I needed and then some. Overall great experience. Thank you, money well spent.
Very helpful! The problem with being an older attorney sometimes is I forget about other potential legal avenues I might pursue. In this case your research reminded me of Restatement of Judgments 2nd which gave me an excellent argument to hopefully defeat collateral estoppel.
This is great and exactly what I was hoping for. As a solo practitioner who has zero staff, this type of assistance is invaluable. Thank you, again!
Learn More!
Interested in learning more about our AI-enhanced Let Us Do Your Research service? Visit our website for more information and to see a sample research memo. https://kcll.org/paid-
research/ We’re ready to work on your research and promise not to take “lazy” short-cuts or sass!
1 Andrew Griffin, ChatGPT Users Complain the AI is Getting Lazy and Sassy, The Independent (Dec 11, 2023)
2 See Emily Price, OpenAI Acknowledges GPT-4 is Getting ‘Lazy’, PCMag (Dec 9, 2023) https://www.pcmag.com/news/openai-acknowledges-
gpt-4-is-getting-lazy
3 See Steve Lohr, A.I. is Coming for Lawyers, Again, New York Times (April 10, 2023)
