A Fast Run Through the First Artificial Intelligence Backlash

When the AI story broke in the spring, most articles expressed amazement, and tried to cover the matter without hype.  Soon thereafter, the possibility of a major AI-caused disaster took over, and people wrote long, hazy pieces speculating on the chance of humanity going extinct under ChatGPT’s virtual boot.  For the next couple of months, AI dominated the technology-related press, to the point where I could not find any material for the regular robotic subsegments on my radio gigs.  It weakened in late May and early June, when commentators seemed to be talked out, and reappeared with a different tone last month, as a combination of reservations, limitations, negative thoughts, and legal reactions, combined with a lack of new AI achievements, took over.  Here I will go through 12 published things from the past eight weeks.

It's a blessing when a powerful group admits they don’t understand something, so it was good to read that “Congress seeks crash course on AI from insiders,” by Cat Zakrzewski and Cristiano Lima, in the June 18^th Washington Post.  Legislators hope to achieve, per Representative Mark Takano of California, “a repository of expertise that is more in an anticipatory mode, that has quicker turnarounds, that can deliver responses more quickly,” and “is not tainted or connected to commercial interests.”  Let us wish them success.

Five days later we read in Engadget.com, “US lawyers fined $5,000 after including fake case citations generated by ChatGPT” (Sarah Fielding).  AI laziness is already a phenomenon, and this kind of result – both errors and a judge discovering them – could sink the technology for many valuable-seeming purposes.  As people have fooled driverless cars through their data-collecting systems, they could large language models, by publishing bogus legal precedents and other “factual” accounts of things which did not happen.  Confusing reality with fiction is also why we need to see “How AI will revolutionize politics in 2024, and why voters must be vigilant” (Brian Athey, Fox News, June 2^nd).  Beyond ability to generate high-quality videos and other campaign tools, it can produce “misinformation, deep fakes and… imagery.”  As we learned in the past two elections about voters absorbing untruths, we can see that there will be plenty more next year, some of which may be blocked, as “A.I.’s Use in Elections Sets Off a Scramble for Guardrails” (Tiffany Hsu and Steven Lee Myers, The New York Times, June 25^th). 

In five days there were three actions taken against the technology.  Per Riddhi Setty in Bloomberg Law on July 10^th, “Sarah Silverman, Authors Hit OpenAI, Meta With Copyright Suits.”  We’re in the early stages of automata not only absorbing proprietary material but circulating it uncredited.  These class-action efforts were filed by three authors, and could become landmarks.  In The Economist’s July 14^th “The World in Brief,” we saw that “America’s Federal Trade Commission opened an inquiry into OpenAI’s handling of consumer data and security practices… reportedly probing whether the startup’s products – principally ChatGPT – could cause reputational damage by publishing misleading information about real people,” and “that training artificial intelligence on personal information could be a form of fraud.”  The same source reported that the Hollywood writers and actors now on strike, among other things, are “also asking for guarantees that AI will not be used to replace them,” which may be the first time surging unions took on burgeoning technology but surely won’t be the last.

The common thread here is AI’s use of incoming data.  That has recently gone further elsewhere, as “’Not for Machines to Harvest’: Data Revolts Break Out Against A.I.” (Sheera Frenkel and Stuart A. Thompson, July 15^th, The New York Times.)  The lead example is from an author who discovered that “a data company had copied her stories and fed them into the artificial intelligence technology underlying ChatGPT.”  Nothing should be controversial here – for decades, routine copyright notices have proscribed putting such content into electronic banks – and this, if anything, is worse.  Expanding even more is “Big Tech took your data to train AI.  We’re suing them for it” (Ryan Clarkson, Fox News, July 19^th).  It’s no certainty that companies can use personal information without clear permission.  Overall, we’re heading for a huge legal collision, and it doesn’t look good for AI.

One odd thing about AI has been its practitioners asking for more regulation.  That doesn’t seem like a way of getting a competitive advantage, such as McDonald’s supporting higher minimum wages knowing they can withstand them better than their competitors, but shows not only uncertainty about its safety but also the tendency of experts closer to a field being more pessimistic than general prognosticators, as documented in “Bringing down the curtain” (The Economist, July 15^th). That may account for the newly visible AI startup, Anthropic, (“Inside the White-Hot Center of A.I. Doomerism,” Kevin Roose, The New York Times, July 11^th), with huge office signs saying THINK SAFETY on display, workers talking freely about their products’ severe dangers, and many reading about nuclear bomb history and comparing “themselves to modern-day (bomb-inventing) Robert Oppenheimers.” 

Yet the last word, for now, goes to something very American.  In “A Blessing and a Boogeyman: Advertisers Warily Embrace A.I.” (July 18^th, The New York Times), in which authors Tiffany Hsu and Yiwen Lu related how products are mentioning the technology in their pitches, from the aforementioned restaurant chain citing a ChatGPT endorsement-of-sorts to “digital avatars” of famous people hawking products.  There will be much more – there need be no existential uncertainty about that.

I will have no post next week.  I will return with the jobs report and AJSN two weeks from this morning.

Employee Choices: Work Tasks, Life Balance, Career Portfolios, Gigs, Other Side Hustles

Workers have not only different choices to improve their positions, but more than before.  Beyond the permanent possibility of just changing jobs, there are others with various merits and considerations.

One is during business hours, the idea that people should “Learn to Say “No” to Extra Work” (Jane Wells, Jane Wells Bulletin, August 29^th).  It is geared toward women, with the author claiming that they get most of the unrewarding assignments.  Whether or not that is true, the issue is real.  Most cubicle-job workers have opportunities to take on and decline tasks, more than many realize.  Suggestions the author cited from a book on this subject were not to accept assignments that cannot lead to promotions, not to join a “board or committee (At Least Not Long-Term),” not to respond at all to emails offering such tasks, and, when such is offered at a meeting, to “mimic the behavior of… colleagues who have no plan to raise their hands by shuffling papers, getting up to leave, refusing to make eye contact.”  Anyone who sees themselves doing the opposites of these should consider changing tactics.  The challenge is that career-strong assignments are often not easy to identify or, especially, to obtain.

An old expression brought to the fore by our chaotic remote-work situation is the subject of “Work-life balance matters to employees – how to find out if it matters to a company” (Lee Hafner, Employee Benefits, January 5).  Per a FlexJobs career services manager, one way is to ask, at interview time, “What would a typical work week look like?”  Another to ask is “how employees are encouraged to have a work-life balance.”  Job seekers should also determine if the company has “a strong onboarding plan.”  All valuable, but it is necessary for everyone to know specifically what “work-life balance” means personally to them.

In Yahoo News on January 3^rd, Scott Sonenshein told us that “Americans are taking more control over their work lives – because they have to.”  As the author put it, “the pandemic accelerated a development that began years ago when workers realized they needed to take on more responsibility for directing their careers,” which led to “career portfolioing,” or assembling and maintaining multiple income.  sources.  That is nothing new in, to name two countries, Iceland and the Bahamas, but it is not traditionally American.  The largest advantage beyond sheer money is that “when facing difficult times at one job, people can turn to other parts of their career portfolio for security and stability.”  If, indeed, you see “a future in which uncertainty is too high to rely on a single institution to fulfill basic needs,” expanding into a portfolio may be for you.

One common addition to such income groupings comes with its own hazards.  To deal with them, Lee Hafner’s May 25^th Benefits News piece, “4 ways gig workers can protect themselves from scams and lawsuits,” may help.  His suggestions are to “choose the right insurance,” as gig workers are legally vulnerable independent contractors, “consult an insurance expert” especially one on the kind of gig work you want to do, “invest in strong cybersecurity” to avoid being responsible for stolen material, and “make a contract, check it twice,” with emphasis on minimizing liability.  These precautions may be more valuable for tasks like the wood drilling pictured with the article than for others, but are worthy of consideration.

Finally, another old thing is becoming new to some in “Beyond ‘Quiet Quitting’: Another Workplace Trend is Making Employers Even Angrier” (Veronika Bondarenko, The Street, May 30^th).  According to “a wide-scale study conducted by consulting firm Deloitte, 46% of polled Generation Z workers and 37% of millennials said that they worked a second part-time or even full-time job in tandem to their main work.”  Not all of these were related to their primary field, as “some of the most popular side hustles include selling online products, delivering food orders or working for a ride-share company and writing marketing materials.”  Per related Bankrate research, such endeavors are “often a way for low-earning employees to keep up with the cost of living,” with the money it brings in “essential.”  Working two jobs to make ends meet is nothing new, and employers, if it does not involve using company resources, should tolerate it.  Or they could just pay more.  As coarse as that seems, it would alleviate or totally remove many of the issues here.  Choices on these issues are not, after all, always made by workers. 

June Jobs Report: Both Seeking and Hiring Vigorous, AJSN Says Latent Demand Up 400,000

This morning’s Bureau of Labor Statistics Employment Situation Summary was supposed to be especially important, especially for investments and interest rates.  So what happened?

At 209,000, the number of net new nonfarm payroll positions fell a sliver short of a 225,000 prediction.  Adjusted unemployment lost 0.1% to reach 3.6%, with the unadjusted variety up 0.4%, seasonally, to 3.8%.  Total employed increased 557,000 to 161,559,000, with the number of people not interested down 867,000, on top of last month’s 1.15 million, to just over 93 million.  The count of unemployed, measured seasonally, was off 100,000 but actually rose 650,000. 

While the number of those officially jobless for 27 weeks or longer, at 1.1 million, was down 100,000, the count working part-time for economic reasons, or keeping shorter-hours positions while seeking full-time ones, soared over 400,000 and is now 4.2 million.  The labor force participation rate was 62.6% for the fourth straight month, and the employment-population ratio sat at 60.3 percent.  Average hourly private nonfarm payroll earnings increased 12 cents, more than inflation, and are now $33.58.

The American Job Shortage Number or AJSN, the measure showing how many more opportunities could be quickly filled if all knew they would be easy and routine to get, increased 380,000 to the following:

 

People with the statuses of marginal attachment, the second through eighth rows here, sharply decreased in number, with the three largest changes reducing the AJSN.  The share of the metric from those formally unemployed increased 2.8% to 34.1%, meaning that almost two thirds of those not working who would accept a suitable but now unadvertised position have another status.  Compared with a year before, the AJSN has lost 268,000, most of which was from fewer people wanting work but not looking for it during the past year.

What happened this time?  There are ever more jobs, more people wanting them, more getting and keeping them, and more people seeking more hours.  Despite our aging population, the number of people not in the labor force fell by one million.  Latent demand increased again, despite net new positions outstripping our population rises, as additional people – two million more than in April – said they are interested in work.  The employment situation is dynamic all around, and that is nothing to take for granted.  The turtle, once again, took a robust step forward.

Five Good Shots Against Remote Work

Now that the pandemic has eased into an endemic, if that, the medical need for people to work from home has disappeared.  Yet one thing it told us is that many employees would like to do that.  Is remote work a good idea?  Here are several pieces maintaining it is not.

In “You Call This ‘Flexible Work’?,” in the New York Times on April 12^th, Fred Turner contrasted the current situation with the 1938 establishment of the Fair Labor Standards Act, which “formally ratified the division of work time from free time.”  Per Turner, “until recently, the physical distance between workplace and home helped guarantee those terms.  The commute enforced a boundary between professional and personal time that millions observed every day.  So, too, did the calendar, dividing days into weekdays for work, weekends for leisure.”  Since then, people have spent less time on commuting, but in the past two decades our homes’ “walls had been well and truly breached,” as “everything we do online can be tracked,” and employers can even “peer into our living rooms, learn a great deal about who we are and use it to alter the terms of our employment.”  And some wonder why unions have made a comeback.

One chronic problem with employment beyond offices is “What Young Workers Miss Without the ‘Power of Proximity’” (Emma Goldberg and Ben Casselman, The New York Times, April 24^th).  Both formal studies and common knowledge have told us that ample feedback is not only possible but achieved mostly when people are physically near their bosses or mentors.  The problem of remote workers being “out of sight, out of mind,” has not been solved, and neither has the downside of remote meetings. 

There are plenty of objectors to people working from home on the other side of the desk, as “Bosses are fed up with remote work for 4 main reasons.  Some of them are undeniable” (Jane Thier, Fortune, June 14^th).  “The golden age of remote work seems to be ending,” as there is “increasing anti-remote literature” and “even tech firms (the first industry that told employees they could work from home forever just a few years ago) are getting engineers and project managers back in the office.”  Thier’s four reasons are “remote work is bad for new hires and junior employees,” “workers admit that remote work (sometimes) causes more problems than in-person work” (with unaligned office days), “remote workers put in 3.5 hours less per week compared to in-person workers” documented in a 2022 Liberty Street Economics report, and “productivity plummets on days when everyone is working remotely (anecdotally)” – especially on Friday afternoons?  In all, “the tide is turning.”

Though how individual workers manage it varies, working from home provides a broader, richer, and more satisfying set of goofing-off opportunities.  Alyssa Place told us how many protect themselves from inquiries into them in “Caught!  Remote employees reveal their top excuses for not working,” on April 20^th in Employee Benefit News.  They were “technical difficulties,” “family or personal emergencies,” “illness,” “misunderstandings,” “distractions and interruptions,” and “other work obligations.”  These all can be legitimate, but the same ones over and over, especially from the same workers, can be telling.

Perhaps pithiest, and therefore most scathing, was “The working-from-home illusion fades,” subtitled “It is not more productive than being in an office, after all,” on June 28^th in the “Free exchange” column in The Economist.  As “a gradual reverse migration is under way, from Zoom to the conference room,” “new research,” including a paper showing that workers handled fewer calls with less efficiency when working from home, has shown that “offices, for all their flaws, remain essential.”  As a result, “higher productivity” will direct supervisors to some combination of office mandates and lower pay for remote-only positions.

The pendulum between work from home and work from offices, as it has since the 1990s, is moving back and forth.  Its motion was disturbed by Covid-19, but the spirit of present times is toward the latter.  There are other sides to this controversy, but for now, awaiting a possible 2030s rediscovery of the advantages of working from home, the office is winning. 

Artificial Intelligence – Key Issues and Considerations – III

Continuing on…

In some technical boom times, the companies profiting most have been those providing supporting services.  In The Economist’s The Bottom Line newsletter on June 3^rd, Guy Scriven looked at that for AI in “Selling shovels in a gold rush.”  He named Amazon Web Services and Microsoft Azure for storage, Digital Realty and Equinix for entire data centers, Wistron and Inventec which “assemble servers for the cloud giants and for local data centres,” and “80-odd” more firms with other services. 

We know something about which positions won’t be affected much by AI, but how about those in the front lines?  Per Aaron Mok and Jacob Zinkula in Insider on June 4^th, “ChatGPT may be coming for our jobs.  Here are the 10 roles that AI is most likely to replace.”  First mentioned is “tech jobs,” namely “coders, computer programmers, software engineers, and data analysts.”  I wrote in 2012’s Work’s New Age that these positions were unusually susceptible to automation, got little published agreement elsewhere, and now that may finally come true.  Others were “media jobs (advertising, content creation, technical writing, journalism),” “legal industry jobs (paralegals, legal assistants),” market research analysts, teachers, “finance jobs (financial analysts, personal financial advisors),” traders, graphic designers, accountants, and customer service agents.  It is easy to see that all will be replaceable by what AI and other electronic services can offer, but as we saw before, embedded worker bases are resistant.

One view bound to get attention is “Big Tech Is Bad.  Big A.I. Will Be Worse” (Daron Acemoglu and Simon Johnson, The New York Times, June 9^th).  With billions of people benefiting from these products, I don’t care for that premise, but if you substitute, say, “dominating” for the adjective, it makes sense.  It’s unavoidable, though, as such nonphysical and portable things are natural monopoly or oligopoly fields, and regulation will be developed along with, or at least soon after, its proliferation.

We reached a landmark earlier this month.  Per Bloomberg Technology’s “This Week in AI” on June 10^th, “ChatGPT creator OpenAI was hit with its first defamation lawsuit over its chatbot’s hallucinations,” as “a Georgia radio host says it made up a legal complaint accusing him of embezzling money.”  The suit is strong, and we’re about to get some precedents – nonfictional ones – on this issue that could soon become depressingly commonplace.

Cade Metz asked, in the June 10^th New York Times, “How Could A.I. Destroy Humanity?”.  The problem seems to center around autonomy, especially if such systems were allowed access into “vital infrastructure, including power grids, stock markets and military weapons.”  Though still limited and not successful, “researchers are transforming chatbots like ChatGPT into systems that can take actions based on the text they generate.”  Given goals, such software will do anything it can to achieve them, for example, “researchers recently showed that one system was able to hire a human online to defeat a Captcha test.  When the human asked if it was “a robot,” the system lied and said it was a person with a visual impairment.”  We will end up blocking off the pathways to true action, but if there are gaps, automata will find them.

Writing about “the Singularity,” or “the moment when a new technology… would unite human and machine, probably for the better but possibly for the worse,” an idea originated by computer scientist John von Neumann in the 1950s, has made a sharp comeback.  And now, we have “Silicon Valley Confronts the Idea That the ‘Singularity’ Is Here” (David Streitfeld, The New York Times, June 11^th).  As AI “is roiling tech, business and politics like nothing in recent memory,” resulting in “extravagant claims and wild assertions,” some think that massive transition is at hand or nearly so.  One long-time advocate, author and inventor Ray Kurzweil, now forecasts it to arrive by the 2040s, but “critics counter that even the impressive results of (large language models) are a far cry from the enormous, global intelligence long promised by the Singularity.”  So we will see, but not today or tomorrow.

Finally, back to the counting-house.  Per Yiwen Lu, on June 14^th and also in the Times, “Generative A.I. Can Add $4.4 Trillion in Value to Global Economy, Study Says.”  I’ve seen a lot of trillions in the news lately, especially in American deficits and capitalization of the hugest companies, and here is another.  Per this McKinsey effort, this one is annually, but “up to,” as “the vast majority of generative A.I.’s economic value will most likely come from helping workers automate tasks in customer operations, sales, software engineering, and research and development” – mostly consistent with the Insider article above.  Just one trillion dollars is $1,000,000,000,000 – how long will it be before we start talking about quadrillions?  And what will the status of artificial intelligence be then?

Artificial Intelligence – Key Issues and Considerations – II

What’s been happening with AI?  This time nothing new technically, but a host of preparations for then.

While not really “The rise of the chatbots” (Rachel Metz, Bloomberg Tech Daily, May 25^th), maybe we can say “It’s raining chatbots,” as “there are AI chatbots in drive-thrus.  They’ve been built into Snapchat.  They’re recommending recipes at BuzzFeed and, disturbingly, have replaced human assistance at the National Eating Disorders Association.“  But this piece was most interesting for the money being raised and assigned to them:  $450 million for Anthropic “in its last funding round” bringing it to “more than $1 billion thus far,” $150 million for Character.AI, and $101 million for Stability AI, all hoping to change the current situation in which “none of these contenders has so far appeared to rival ChatGPT in terms of consumer popularity, name recognition, or funding” – even the latter.

Are we really looking on as “A Hiring Law Blazes a Path for A.I. Regulation” (Steve Lohr, The New York Times, May 25^th)?  Well, although there have been AI regulations in place since at least 2021, every new one can be a meaningful precedent.  Now, in New York, “the city’s law requires companies using A.I. software in hiring to notify candidates that an automated system is being used.  It also requires companies to have independent auditors check the technology annually for bias.  Candidates can request and be told what data is being collected and analyzed.  Companies will be fined for violations.”  How this law is enforced, what firms and jobseekers say about it, and how often it will be broken will all get nationwide attention.

In a related area, “AI is here to stay; it’s time to update your HR policies.” (Breck Sumas, Fox Business, May 27^th).  Organizations, per the owner of a human resources firm, will need to decide which products workers can use on the job, and for what, keeping in mind AI’s great utility and insidious data insecurity, and should be starting to develop, document, and implement those rules now.

Of the people with great AI fears, CEOs of major AI companies are at the top.  While such views are controversial, it may have been surprising for us to see that “A.I. Poses ‘Risk of Extinction,’ Industry Leaders Warn” (Kevin Roose, The New York Times, May 30^th).  The statement, released that day by the nonprofit Center for A.I. Safety and “signed by more than 350 executives, researchers and engineers working in A.I.” including the tops of OpenAI, Google Deep Mind, and Anthropic, read, in full, “Mitigating the risk of extinction from A.I. should be a global priority alongside other societal-scale risks, such as pandemics and nuclear war.”  That organization did not venture a view on how that catastrophe would happen, but the danger of autonomous goal-seeking programs has long been understood.  There are now large groups on both sides of the fear/no fear divide, and arguments between them may not always be harmless and polite.

It's not too early to look at “How AI will revolutionize politics in 2024, and why voters must be vigilant” (Brian Athey, Fox News, June 2^nd).  Although much in AI will change over the next year-plus, we now must consider the difficulty we will have, given excellent-quality synthesized images and falsely attributed written statements, in “discerning reality.”  As well, “copywriting for fundraising emails, captions for social media posts, and scripts for campaign videos can now all be produced with an unprecedented level of speed, personalization, and diversity.”  All of these “are currently being navigated by people whose mandate is to win at all costs,” making ethical behavior sporadic at best.  The past two presidential elections told us a great deal about voters’ often tenuous perception of the truth, and the next may be vastly worse.

Finally, in an area adjoining AI, we see that “Robots could go full ‘Terminator’ after scientists create realistic, self-healing skin” (Emma Colton, Fox News, also on June 2^nd).  Remember the passage in one of those films when someone warned that automata could then have “bad breath”?  Now has been developed “layers of synthetic skin that can now self-recognize and align with each other when injured, simultaneously allowing the skin to continue functioning while healing.”  We may get to interact with people who aren’t people in person, unawares, as well as electronically.  From there… who knows?

Back for more with the next post.

Artificial Intelligence – Key Issues and Considerations – I

Although the next round of AI’s technological progress will be in the background for a while, there’s no getting away from this topic.  It’s cheek-to-jowl with jobs and the economy, and we know little more about what effect it will ultimately have than we knew about that of cars when Karl Benz and Gottlieb Daimler were tinkering with contraptions.  I intend to provide only the most important concerns and leave off the seemingly endless pieces speculating on whether AI will be a boon to or the end of humanity, for the same reasons baseball writer Bill James said about another issue decades ago: “1.  I don’t know, and 2.  You don’t know, either.”  This is the first of at least three such consecutive posts.

Oldest, but still within the month, is “8 Questions About Using AI Responsibly, Answered” (Tsedal Neeley, Harvard Business Review, May 9^th).  After “How should I prepare to introduce AI at my organization?” (“Ensure that everyone has a basic understanding of how digital systems work…  make sure your organization is prepared for continuous adaptation and change… build AI into your operating model”), we got “How can we ensure transparency in how AI makes decisions?” (“Recognize that AI Is invisible and inscrutable and be transparent in presenting and using AI systems… prioritize explanation as a central design goal”), “How can we erect guardrails around LLMs [large language models] so that their responses are true and consistent with the brand image we want to project?” (“Tailor data for appropriate outputs… document data”), “How can we ensure that the dataset we use to train AI models is representative and doesn’t include harmful biases?” (“consider the trade-offs you make…” get “diverse teams” to “collect and produce the data used to train models”), “What are the potential risks of data privacy violations with AI?” (follow the seven Privacy by Design principles), “How can we encourage employees to use AI for productivity purposes and not simply to take shortcuts?” (“evaluate whether AI’s strengths match up to a task and proceed accordingly”), “How worried should we be that AI will replace jobs?” (not across the board), and “How can my organization ensure that the AI we develop or use won’t harm individuals or groups or violate human rights?” (“Slow down and document AI development… establish and protect AI ethics watchdogs… watch where regulation is headed”).  A worthwhile primer.

Four days later, The Economist covered the second-to-last point above in “Your new colleague; Artificial intelligence is about to turn the economy upside down.  Right?”  This article cited a Goldman Sachs paper projecting that “in a best-case scenario generative AI could add about $430 billion to annual global enterprise-software revenues” as 1.1 billion world office workers could require just under $400 each.  Yet it could be slow, considering examples such as a 90-year lag between automating technology and job decimation of telephone operators and the continuing presence of subway-train drivers and traffic police.  Additionally, “it is even possible that the AI economy could become less productive,” as may be the case with smartphones and remote work and certainly was, for a long time, with personal computers. 

Here’s a foundation for something going wrong: “AI tools being used by police who ‘do not understand how these technologies work’: Study” (Chris Eberhart, Fox News, May 15^th).  Respondents were “not familiar with AI, or with the limitations of AI technologies,” although they liked having this capability.  Perhaps a basic course of some sort should be required.

A fine semi-philosophical question hit the press in “Is It Too Late to Regulate A.I., or Too Soon?” (Timothy B. Lee, Slate, May 18^th).  It started with an account of OpenAI CEO Sam Altman’s May 16^th “appearance before the Senate Judiciary Committee” in which the corporate leader asked for licensing “any effort above a certain scale of capabilities, and could take that license away and ensure compliance with safety standards,” with special concern with systems that could “self-replicate and self-exfiltrate into the wild.”  Such a system is being built in Europe.  Either could call for somewhere between scrutiny and a ban on incremental improvements to existing releases, such as ChatGPT4, and could greatly delay availability of future ones.  In the meantime, regulating bodies would need to understand the current issues and technical state, which could also take a while.  No, it’s not too late, but if governments, not noted for being nimble, cannot keep up, it will be too soon.

After all these high-level AI concerns, how about some pithy advice on how to use it?  We got that in “On Tech A.I.: Get the best from ChatGPT with these golden prompts” (Brian X. Chen, The New York Times, May 25^th).  Suggestions here are geared also to Bing, from Microsoft, and Bard, a Google product.  First is that “if you’re concerned about privacy, leave out personal details like your name and where you work,” as it could be shared, omit “trade secrets or sensitive information,” and to be aware that it may have “hallucinations,” as the tools “can make things up” “while trying to predict patterns from their vast training data,” some of which is “wrong.”  From there, use prompts starting with “act as if” continuing with the role you want the software to play and “tell me what else you need to do this.”  Instead of starting fresh, “keep several threads of conversations open and add to them over time.” 

More next week, as this area continues to evolve.