"Believe all women" was always a bad slogan, even if taken in context. There are women who make a living from lying, such as actresses, comediennes, novelists... lawyers!
There is now a very kind name --AI Hallucinations-- for when real life lawyers use AI for their research and trust AI so much that they skip verification. Apparently, there are now hundreds of cases where lawyers in court cited legal precedents that never existed in real life.
One lawyer in New York submitted a brief containing six AI fabricated cases; two attorneys in the Sixth Circuit wrote up over twenty-four "hallucinated" case citations; in California, an attorney was fined $10,000 for using twenty-one fake cases (generated by AI) in support of an appeal.
As a result of severe rebukes and hefty fines, the legal community is now wary of using Chat without verification, and someone has set up a website listing all known AI Hallucinations.
Presumably, the tech wizards who trained their AI or ChatGPT on large libraries of fiction (or pirate sites) should not have included entire genres, such as legal thrillers.
The week before last, I learned that my State put "a new computer system" in place that is generating notices in error relating to tax returns. In my case, I was sent a refund check that --if I had encashed it-- might have paid for three cups of overpriced coffee, and subsequently I received a notice claiming that I had underpaid my State taxes by $10,000. Same State, same return, same tax year.
Both notices cannot be true. In my opinion, neither is accurate.
I was told, don't take the $28 refund (but keep it safe). Prove with a jpg of the cancelled check that you paid your taxes.
The watchword for use of AI should be, "Trust But Verify". One cannot "Believe all AI".
One assumes that AI can be highly beneficial in improving weather forecasting, although perhaps not so much regarding "climate", and in making medical research and diagnoses more efficient, and in myriad ways including warfare, traffic control, space exploration, and crime solving.
AI might be a tad worrisome when it comes to the privacy invasions and possibilites for error in age gating, in surveillance and facial recognition (false positives), and in deep fakes, and in encryption and decryption.
Dr. Matthew Guariglia is the Senior Policy for EFF (the Electronic Freedom Foundation). Last week he testified to the House Homeland Security Subcommittee on Cybersecurity and Infrastructure Protection.
I wonder if there is an acronym for that. EFF's biggest concerns are that AI-powered mass surveillance violated constitutional rights and protections in a big way, and that government secrecy prevents the public and lawmakers from knowing when AI makes mistakes.
As we see, AI does make mistakes.
EFF on privacy https://www.eff.org/issues/privacy
EFF on AI https://www.eff.org/issues/ai
Meanwhile, Apple automatically saves user IDs and passwords not only to the primary device but also to linked devices. One might not notice, but maybe it is not a good idea to have banking logins saved and shared, because if it goes to various synched devices, it probably gets stored on a server farm.
They may not be as safe as you think.
Not if Q-Day is as nightmarish as some think it will be.
Paloalto Networks explains: https://www.paloaltonetworks.com/cyberpedia/what-is-q-day
In a nutshell, Q-Day is the day when a quantum computer becomes capable of breaking the encryption standards that underpin modern digital life. All passwords, banking, cloud storage, Bitcoins, everything.
All this, at the same time that credit card companies, banks, insurance companies, periodicals etc put pressure on clients and customers to agree to "go digital" and avoid the waste and insecurity of paper statements and tax forms in the mail.
One might want to preserve some hard copies, just in case Q-Day is upon us.
Hence, FUBAR.
All the best,
Rowena Cherry
Posts
No comments:
Post a Comment