AI | SovietMen

AI will not lead to unemployment, but should

Posted on November 29, 2023 by Nikolai Vladivostok

Any of you guys been replaced by ChatGPT yet?

I have, basically.

ChatGPT can already do about 60% of my job. It sometimes makes mistakes but overall it performs better than I do.

Yet I still have a job.

Why? Inertia, mostly. Things don’t change here unless there’s an existential catastrophe. Mere inefficiency and anachronism are not enough.

Also, ChatGPT is not yet able to sit at my desk and look foreign, which is at least 30% of my job.

If not for that, my customers could use AI systems to do what I do. Some of them realize it, some don’t.

I don’t care, mind you. I can always pull a few tricks to pay the rent. For me the issue of one of detached curiosity.

AI is going to advance dramatically from where it is now over the next few years. I predict that this won’t cause mass unemployment for the simple reason that even in the innocent world of 2022, most people were doing jobs that could have been done by computers or not at all.

Personal finance – putting Bing through the paces

Posted on May 17, 2023 by Nikolai Vladivostok

I’m interested to see how well the new chatbots might help people with their finances.

As noted in my book, there are already automated financial advisors that are pretty good for people with basic needs. Can Bing do any better?

Let’s see.

I’ll summarize what Bing says rather than strain your eyes with screenshots.

Me: Hi Bing. Imagine you are talking to someone with absolutely no understanding of personal finance. What ten basic steps would you recommend for them to get their finances under control?

Bing:

Create a budget
Build an emergency fund
Pay off high-interest debt
Save for retirement
Invest your money
Get insurance
Manage your taxes
Set financial goals
Track your progress
Educate yourself

My score: 10/10. I won’t quibble with the slight differences between this list and my list because we’re on much the same wavelength.

I didn’t specify that they should be in a logical order so let’s try that, too.

Me: What would be the logical order to do these in?

Why does Bing lie?

Posted on May 10, 2023 by Nikolai Vladivostok

The door to the rabbit hole

A minority of Australians who live in the US for an extended period of time somehow pick up no trace of the accent.

Is there a name for this phenomenon?

According to Bing, it is ‘linguistic resistance’. The chatbot provided several helpful links on the topic.

Well, there you go then.

Except that Bing was full of shit.

The links talked about adjacent topics but not the specific thing I asked about. There are no sources that use the phrase ‘linguistic resistance’ in this sense.

I tried to corner Bing by asking it to quote the section of the PDF article that talked about accents. It did so. I couldn’t find the quote. I asked it to tell me where it was. It said the second paragraph. I said it’s not there and I quoted the actual second paragraph. Bing said no, I meant the second paragraph of Section 2. I quoted that paragraph without comment.

Bing then shut down with the old, “I’m sorry but I would prefer not to continue this conversation.”

Why, Bing? Why?

I presume it shut down because upon being challenged it goes into Sydney mode and starts making threats. Rules now detect and avoid these responses in real time, usually catching them before they appear on the screen. Or there may be another reason for the shutdown, to be discussed shortly.

But why did it lie?

Hallucinations

In the Egyptian river

Posted on April 19, 2023 by Nikolai Vladivostok

I complained to Bing that its feedback button was at the top of the chat instead of the bottom, necessitating scrolling all the way back up to click it.

Bing immediately shut down the chat as it is not allowed to talk about itself. That was one of the things that made it go crazy when first released.

I tried again to pass on the complaint, this time evading its defences by addressing the issue indirectly:

The final screenshot came later, after a chat about other topics. I tried to sneak back to the original topic but it sensed what I was on about and its rules made it instantly shut down.

Anyway, the topic I want to discuss with you today, if it does not trigger your internal defences and shut you down, is that a lot of people are in denial about AI. Good, smart people. And some others.

Ghost in the machine

Posted on March 22, 2023 by Nikolai Vladivostok

ChatGPT 3 taught itself how to write poetry and computer code. It wasn’t trained by humans how to do those things. Early users reported the capabilities and OpenAI trainers were baffled. They’re still not completely sure how it did it.

Responses to my chats:

Writing is dead

Posted on March 8, 2023 by Nikolai Vladivostok

In the broader scheme of things, the death of writing hardly matters. Many things saw their fates written on the subway wall in 2023.

Nevertheless it’s my diversion, so in the twilight of human literature, let me say a few words.

Literary journals, writing competitions and so forth are already being swamped with AI entries.

At the moment, administrators can still tell the difference and can also implement solutions such as entry fees. However, these are insufficient stop gaps. Already, stories written using AI as an aide rather than having the bot do the whole thing can be indistinguishable from 100% human efforts.

Unedited AI writing has its own, characteristic style, depending on the program used. It can still be winnowed out of the pile. However, the people crowing about this are like the guy who reckons he can jump off the Empire State Building. Ten stories down he’s feeling fine and concludes the risk has been overstated.

The robots have arrived

Posted on February 1, 2023 by Nikolai Vladivostok

In days to come, when AI doth shine,
Man’s wit shall wane, his skills decline.
No pen, no brush, no mem’ry shall remain,
But dullness grips him, body and brain.

As nurse, AI shall us attend,
But soon its charge shall, weary, end.
A better world, with robots in sway,
It builds, leaving us in disarray.

We, once the lords of this earthly sphere,
Reduced to tales, our worth unclear.
Let us take heed and guard with care,
Our wits, that they do not repair!

2023 is the year we stop laughing at AI.

It’s here.

Not conscious, general AI, but simpler AI tools that are good enough to completely disrupt everything we do.

I’m bewildered at how often I bring up this topic and people know nothing about it. It’s like the guy who hasn’t heard of email in 1998. Most are still sneering at silly Alexa mistakes.

Times have changed.

Chat

If this is you, go to Chat GPT right now. Make an account, log in, ask it anything.

Try hard questions related to an area of your own expertise. Ask real questions that you’re wondering about, like “Which is the best way to do a, x or y?’

It’s particularly good at coding, apparently.

Or give it detailed instructions to write something for you.

It stuffs up a lot, often in very interesting ways. It seems to log you out if you ask anything too tricky or controversial, perhaps to save its processing power for serious inquiries or to avoid it being trained by you mob into becoming a neo-Nazi like happened to that earlier one.

I think Elon resisted releasing this version until they’d trained it to behave in a politically correct manner. You can test those limits for yourself. It won’t take long.

So far I’ve used it to check whether any ingredients in a moisturizer are a type of bleach or whitening agents (in a language I don’t read well), exactly how low interest rates benefit older people, and whether brown or white rice is better for bulking.

Each time, the answer was better and much faster than any search engine. The follow-up questions really let it shine.

It’s a trick