As Good As It Gets is not as good as its Oscar haul would suggest, but as a haunting precautionary warning, its titular line ranks up there with “all glory is fleeting” or “never go in against a Sicilian when death is on the line.” It can be read as a call to forfeit your dreams to live a second-best life, but in the grand scheme it’s more generously an admonition not to allow dreams to transform into anchors dragging you down.
Generative AI stormed into the public consciousness when ChatGPT arrived to ghostwrite every middle school essay and, occasionally, federal court briefs. In less than two years, the technology has improved considerably and — more importantly — with folks building the necessary guardrails and firm tech professionals crafting sound procedures, it’s become a credible tool for the legal industry. It certainly doesn’t do anything well enough to replace massive troves of associates yet. But it summarizes, it streamlines drafting, and with proper guidance, it can perform limited research tasks, which are all use cases with tangible value for legal work.
But while AI gurus preach never-ending advancement… what if this is as good as it gets?
A few weeks ago on the Legaltech Week Journalists’ Roundtable, I cited recent studies suggesting that Generative AI development was sapping electricity — and water for cooling — at an alarming and expensive rate that could become unsustainable. Goldman Sachs seems to agree.
In a new report, the bank compared the mounting costs of GenAI development with the plausible opportunities for future revenue and came up empty. The tasks that GenAI performs now are, the report suggests, likely the only tasks it can ever support. At least to a level capable of challenging, much less replacing, a human. With the applications of GenAI largely capped, Goldman sees the potential revenue streams as… slightly bigger streams and not raging rivers.
Which is a significant problem because GenAI needs a whole lot of money to improve. Few have been as bearish on GenAI as Ed Zitron, and in his victory lap coverage of the Goldman report, he notes that even linear improvements to large language model performance will require exponential increases in training data:
Anthropic CEO Dario Amodei saying that the AI models currently in development will cost as much as $1bn to train, and within three years we may see models that cost as much as “ten or a hundred billion” dollars, or roughly three times the GDP of Estonia.
No one is dumb enough to spend a hundred billion dollars to improve incrementally. Maybe the people running the F-35 program. But outside of the defense industry, no one is dumb enough to spend a hundred billion dollars to improve incrementally.
Maybe this isn’t all that bad for the legal industry. We’ve written before about the importance of rejecting the frame that AI should be more “human” and instead focus on its power to accelerate purely mechanistic tasks. The AI impresarios that Goldman swats down are pitching a Scarlett Johansson sexbot, but legal doesn’t really need that.
Seriously, legal does not need that.
Lawyers aren’t — at least hopefully not for ethics reasons — expecting a robot to make legal judgments for them. They just want a tool that they can feed full of a whole production and then spit out all the key issues it finds based on basic legal concepts. Most of that doesn’t even require the generative part of AI that’s causing all these cost concerns.
If the bottom falls out of GenAI development, then that’s not great for legal. But assuming it trundles along just getting marginally better at what it does already, that’s probably still a good deal for legal.
So… what if this is as good as it gets? Are Silicon Valley techbros going to crash a good thing because they fancy themselves as some post-modern Prometheus or are they going to pull back and develop a nice, usefully boring productivity tool?
Earlier: The Legal Industry Has A Long Way To Go Before GPT Matches The Talk
Maybe We’ve Got The Artificial Intelligence In Law ‘Problem’ All Wrong
Joe Patrice is a senior editor at Above the Law and co-host of Thinking Like A Lawyer. Feel free to email any tips, questions, or comments. Follow him on Twitter if you’re interested in law, politics, and a healthy dose of college sports news. Joe also serves as a Managing Director at RPN Executive Search.
