Interesting article.
This seems like a good approach.
In the meantime, shipping companies could simply be more transparent and notify the captain of any messages sent to a ship including the help of AI. This notification can be as simple as “ChatGPT was used to correct grammar in this email” to “here are some additional ideas to that problem created by Bing AI”.
Chatgpt is going to be the worlds best excuse for just about everything…
Indeed. I’m going to start saying “ChatGPT told me to do that…” instead of “I meant to do that…”
From the article:
At this point, it’s difficult to prove the message was created by an AI, but it’s certainly in a style similar to that produced by AI…
Not sure that’s valid conclusion. ChatGPT can write in any style you direct. For example, here’s a recent NMC announcement that I asked ChatGPT to re-write as rhyming verse written by a pirate:
Ahoy me hearties, listen up ye swabs
We got news to share, no need for jabs
National Maritime Center (NMC), hear ye hear ye
E-signatures be accepted, it’s a decree
CG-719 forms and supporting docs, it’s true
For merchant mariner credentials, medical certs too
E-sign yer applications with software-generated signs
Like Microsoft Word or Adobe Acrobat, they’ll be just fine
Sea service, assessments, training docs and more
Qualified assessor requests, we got ye covered, for sure
But beware me mateys, any missin’ signs
Will be returned to ye, causing delays and whines
Submit yer MMC app to regional exam centers
Or MEDAIP@uscg.mil for medical certs, no need to enter E-mailing yer app with e-signatures be our preference
But check yer e-mail provider, avoid any inconvenience
Remember, me hearties, incomplete info will be rejected
Cause delays and frustration, and leave ye dejected
For any queries, contact NMC’s customer service crew
E-mail IASKNMC@USCG.mil, or call 1-888-IASKNMC, they’ll see ye through
Ye can also scan the QR code, visit the website or use online chat
For assistance and guidance, we got yer back, that’s a fact.
Then, who will be jailed?
The inventor of ChatGTP?
Last night I was listening to a podcast and I heard an advertisement for a ChatGPT-type app, with a big difference:
It is all voice based. You presumably talk to the app for awhile, expressly for it to learn your voice in order to mimic it, and sound just like you. The rest is just a Chat-bot, but with the app speaking in your voice.
Quite terrifying. I hope I wasn’t listening correctly, and it was all parody. But it didn’t sound like that…
It’s here. I saw a bit on it on TV a day or so ago. We’re not that far from being able to order a walking talking robot with our voice and appearance.
Voice is a powerful thing. The human/eye brain is pretty good at sorting out a real face from an animated one, even today. But a voice absent a face can be compelling. I can see this going sideways in a hurry.
who went to prison for the 2008 banking/mortgage CDO scandal?
It’s already being used in scam phone calls from “relatives”.
Alan Turing’s name comes up as AI royalty but he wouldn’t be available for a grilling from Congress. His life is depicted in a movie titled Breaking the Code.
the scam was phone messages saying a relative was in trouble send money , now you will be talking to a robot that sounds like your relative, plus we already had the technology top make a video of your face saying things it has never said…
An email exchange on whether the machines will kill us
From Marginal Revolution
ChatGPT was GPT-3.5 now GPT-4
#dbeierl - Imagine the voice bot’s commercial potential. When the floating billboard boys connect the dots they’ll deliver their messages using your mom and dad’s voices over loudspeakers. In smarmy euphemistic prose, they’ll present it as “an opportunity for the beachgoer to enjoy the interactive canvas experience on a deeper level.”
I see the term “digital watermark” mentioned in context of people using chatbots to create or assist in creating documents such as emails.
If there was communication between parties maybe a checkbox system to show the degree of chatbot use would be useful for, example to show all, some or no assistance was used. Not foolproof of course.
Bing AI describing Silicon Valley Bank collapse in monkey terms
(Memes on Instagram: "Need this for all complex explanations (Via Titkok/build_in_public)")
Great now Captains are going to have to deal with AI emails on the weekend when every shore employee is out of the office. I guess the Monday mornings calls from the charterer after they realize where the vessel is, is now a thing of the past. I love it when you finish cargo and leave port early on a weekend then get calls Monday asking what’s your best speed to X port or can you make it to X port by this date? Etc etc etc. you go from trying to kill time all weekend to racing to the next port…… anyway I’m rambling
I recognize that instructions generated by AI could contain errors. I would say that proper management should catch those errors.
Consider these two scenarios:
If that message gets out of the office without the boss looking at it then whose fault is it if their last deposit goes to S. Crowe’s account?
AI is going to be a tool like writing itself, printing, calculators, spreadsheets, and yes, auto-pilot, electronic charts, and GPS. All those things malfunction at times. At one time in the US large numbers of people freaked out about building elevators without an operator to take them up and down. I’m not making this up. They would walk up the stairs rather than take a chance on an elevator that wasn’t supervised by a person. We’ve mostly gotten over those.
Yuval Harari, in the book “Homo Deus” posits that, in the broad sweep of evolution, we are all just biological algorithms. We are trying to develop an auto that can be safely operated by computer AI. We already have cab drivers who are biological algorithms that do the same thing. You tell the driver you want to go to the Hilton Hotel and they choose a safe route and deliver you there. Of course, the driver might be a novelist working on the great American novel when he isn’t driving, but for you, the cab algorithm works fine.
Do we know that this happened, or is this more Konrad Klickbait?