What will the next half-decade bring? Can we rely on our current impressions of these tools to judge their quality, or will they surprise us with their development?
As someone who has spent many hours playing around with these models, I think many people are in for a shock. LLMs will have significant implications for our business decisions, our portfolios, our regulatory structures and the simple question of how much we as individuals should invest in learning how to use them.
To be clear, I am not an AI sensationalist. I don't think it will lead to mass unemployment, much less the "Skynet goes live" scenario and the resulting destruction of the world. I do think it will prove to be an enduring competitive and learning advantage for the people and institutions able to make use of it.
I have a story for you, about chess and a neural net project called AlphaZero at DeepMind. AlphaZero was set up in late 2017. Almost immediately, it began training by playing hundreds of millions of games of chess against itself. After about four hours, it was the best chess-playing entity that ever had been created. The lesson of this story: Under the right conditions, AI can improve very, very quickly.
LLMs cannot match that pace, as they are dealing with more open and more complex systems, and they also require ongoing corporate investment. Still, the recent advances have been impressive.
I was not wowed by GPT-2, an LLM from 2019. I was intrigued by GPT-3 (2020) and am very impressed by ChatGPT, which is sometimes labeled GPT-3.5 and was released late last year. GPT-4 is on its way, possibly in the first half of this year. In only a few years, these models have gone from being curiosities to being integral to the work routines of many people I know. This semester I'll be teaching my students how to write a paper using LLMs.
ChatGPT, the model released late last year, received a grade of D on an undergraduate labor economics exam given by my colleague Bryan Caplan. Anthropic, a new LLM available in beta form and expected to be released this year, passed our graduate-level law and economics exam with nice, clear answers. (If you're wondering, blind grading was used.) Granted, current results from LLMs are not always impressive. But keep these examples - and that of AlphaZero - in mind. I don't have a prediction for the rate of improvement, but most analogies from the normal economy do not apply. Cars get better by some modest amount each year, as do most other things I buy or use. LLMs, in contrast, can make leaps.
Still, you may be wondering: "What can LLMs do for me?" I have two immediate responses.
First, they can write software code. They do make plenty of mistakes, but it is often easier to edit and correct those mistakes than to write the code from scratch. They also tend to be most useful at writing the boring parts of code, freeing up talented human programmers for experimentation and innovation.
Second, they can be tutors. Such LLMs already exist, and they are going to get much better soon. They can give very interesting answers to questions about almost anything in the human or natural world. They are not always reliable, but they are often useful for new ideas and inspirations, not fact-checking. I expect they will be integrated with fact-checking and search services soon enough. In the meantime, they can improve writing and organize notes.
I've started dividing the people I know into three camps: those who are not yet aware of LLMs; those who complain about their current LLMs; and those who have some inkling of the startling future before us. The intriguing thing about LLMs is that they do not follow smooth, continuous rules of development. Rather they are like a larva due to sprout into a butterfly. It is
only human, if I may use that word, to be anxious about this future. But we should also be ready for it.
(COMMENT, BELOW)
Cowen is a Bloomberg View columnist. He is a professor of economics at George Mason University and writes for the blog Marginal Revolution. His books include "The Complacent Class: The Self-Defeating Quest for the American Dream."
Previously:
• 12/27/22 Beware the dangers of crypto regulation
• 12/27/22 Americans have found their happy place
• 12/14/22 The real risk of higher inflation is lower wages
• 12/07/22 Fight poverty, not income inequality
• 10/10/22 A crisis is coming in Europe. The only question is, which kind?
• 09/06/22 What is the purpose of public policy?
• 08/15/22 The future of travel is less exotic
• 08/01/22 Welcome to the era of antisocial media
• 07/25/22 Biden's COVID diagnosis is a wake-up call for America
• 05/12/22 A nuclear strike might not prompt the reaction you expect
• 03/22/22 Doomscrolling has ruined our sense of time
• 01/22/22 Wokeism has peaked
• 01/31/22 The latest bias to worry about
• 01/17/22 America's loneliness epidemic
• 01/07/22 Some of America's top universities just revealed they're not morally serious
• 12/29/21 America would be more happy with more people
• 12/10/21 Bill Gates, Jeff Bezos, Elon Musk . . . and Paul McCartney
• 12/08/21 The only two pieces of advice you'll ever need
• 11/29/21 Nuclear fusion is close enough to start dreaming
• 10/27/21 America's national mood disorder
• 06/10/21 Lifting of mask mandates poses a challenge for Libertarians
• 05/28/21 Why economics is failing us
• 04/19/21We need green energy. We don't need green jobs
• 04/14/21 Libertarianism isn't dead. It's just reinventing itself
• 04/05/21 What does the world need? More humans
• 02/10/21 If Biden goes big now, he may have to go small later
• 01/12/21 Covid improved how the world does science
• 12/07/20 How to make sure your complaint is heard
• 10/27/20 It's getting better and worse at the same time
• 09/14/20 How to be happy during a pandemic
• 09/04/20 Trump is winning the vaccine debate with public health experts
• 07/01/20 Why Americans are having an emotional reaction to masks
• 05/20/20 Covid-19 will expose the ghosts in the U.S. economy
• 05/07/20 Are aliens visiting us? US military seems to think so
• 05/06/20 America's reopening will depend on one thing --- trust
• 04/22/20 How the covid-19 recession is like World War II
• 04/15/20 America is returning to 1781
• 04/08/20 Covid-19 is is upending everything for status seekers
• 03/17/20 The coronavirus will usher in a new era of entertainment
• 01/28/20 Social Security isn't doomed for younger generations
• 01/08/20 Why 2020 is harder to predict than 2019 was
• 12/02/19 Equality is a mediocre goal so aim for progress
• 11/25/19 Inflation inequality creates winners and losers
• 11/09/19 OK kids. This boomer has had enough
• 10/20/19 Would you bet against Trump in 2020?
• 09/25/19 The right industrial policy for America
• 09/24/19 Harvard's legacies are nothing to be proud of
• 09/02/19 Yes, the Fed could still stop a recession
• 08/20/19 A trade deal with China wouldn't change much
• 07/29/19 How your personality traits affect your paycheck
• 07/16/19 Internet 101 should be a required class
• 05/28/19 How Dems actually are the ANTI-immigrant party
• 04/23/19 Want to help fight climate change? Have more children
• 03/22/19 America isn't as divided as it looks
• 03/12/19 The Twitter takeover of politics: You ain't seen nothing yet
• 03/04/19 How to tell which Dem dreams won't come true
• 02/07/19: Now the Dems want to end America's nuclear first strike option. How clueless is that?
• 01/29/19: The shutdown hit a lot of government workers --- hard. But, ultimately, who is responsible for their unfortunate circumstances?
• 12/12/18: The West is abusing its legal power to punish people or institutions that do things it doesn't like. It better stop
• 10/23/18: The US needs Saudi Arabia, and vice versa
• 10/19/18: The right finds the perfect weapon against the left
• 07/24/18: The drive for the perfect child gets a little scary
• 06/04/18: Side effects of the decline of men in labor market
• 05/14/18: Proving Marx's theories right
• 05/08/18: Holding up a mirror to intellectuals of the left
• 05/01/18: Virtual reality will make lives better ... mostly
• 04/16/18: It's hard to burst your political filter bubbleIt's hard to burst your political filter bubble
• 04/09/18: The missing key to grasping why American politics seems to have become more polarized, with no apparent end in sight
• 04/05/18: Two American power centers are about to clash
• 03/22/18: We fear what we can't control about Uber and Facebook
• 03/08/18: How to stop the licen$ing insanity
• 01/10/18: Polarized Congress needs to bring back earmarks
• 12/27/17: The year when the Internet collides with reality
• 11/07/17: Would you blame the phone for Russian interference?
• 10/23/17: North Korea is playing a longer game than the US
• 10/12/17: Why conservatives should celebrate Thaler's Nobel
• 08/02/17: Too many of today's innovations are focused on solving problems rather than creating something new