eevBLAB 106 - ChatGPT AI Has Changed EVERYTHING

[bob808]

You're not far whatever if you only outdo the most extreme historical ones. You're far any side if you stray too far from the middle, from common sense. Absolutely ok to call out stupidity on both sides.
Having accessible healthcare is not a stupid. How and why would you frame it like that? "Oh I'm far left? So what, you want people to just die without affording medical bills?"

[vk6zgo]

You're not far whatever if you only outdo the most extreme historical ones. You're far any side if you stray too far from the middle, from common sense. Absolutely ok to call out stupidity on both sides.
Having accessible healthcare is not a stupid. How and why would you frame it like that? "Oh I'm far left? So what, you want people to just die without affording medical bills?"

The problem is that there are many on the right side of politics who think exactly like that, & my point is that they are always ready to throw around the currently popular "swearwords" of that group.
At the moment, we are seeing ridiculous, over the top, outrage from people about things, of which whilst some may indeed be described as "stupid", are not the existential threat to society that they are being portrayed as.

[adx]

Yeah I remembered there's this meme that represents the concept.

I see that meme as a mirror image, with the right having run further and further away from centre, dragging the centre with them in some sort neoliberal panic (blinkers on race to the end of infinite consumerist growth, convenient denial of inequality etc). Ok maybe recently the pendulum has been swinging back, carrying all its baggage at once with it - but there are quantifiably obvious reasons for the reaction.

The real problem with centrists is that they exist in this echo chamber they call the "news" and think in terms of concepts like "society", "authority", "money", "science", "economy", "growth", "progress" and so on as if they are real. They add shifting mass to the pendulum, a predictable instability. They think they are correct, they "know" - in common with the extreme left, and right.

Ok, maybe I jest just a wee bit, but I'm becoming more convinced with every passing minute that humanity requires these AI overlords to continue its current path without a particularly sticky end. It doesn't matter whether we think it is sentient, it just has to be better than a democratic majority.

(Right, where's that /sermon tag on the toolbar...)

[dietert1]

Talking about democracy and parties appears a bit oldfashioned. Recent news was that Microsoft increased its share in OpenAI by 2 billion US$. Thiel and/or Musk will do the same if they think they may have a future impact.
Here in Brasil people no longer ask about left versus right but they go after the money. Who paid for the January 8th attacks? For example they found that many of those busses each carrying 30 or 40 people to Brasilia from all over the country were hired for the price of one ticket..

Regards, Dieter

[bob808]

Both extremes constantly pull to each side, it is when centrists decide to move that things actually change, centrists define normality in the end. Hence why they are a juicy target, especially for the left as I've personally noticed. 
Centrists shouldn't average out the craziness on both sides, they should pick the useful things that make sense, from the extremes.
ChatGPT and similar tools have great potential for change, and hence shouldn't be taken over by any of the extremes or things can go sideways really fast. In the end, bottom line, it is always about money and power/control, not the good of society (for the extremes).

[bob808]

I see that meme as a mirror image, with the right having run further and further away from centre,

This meme does not show the extreme left or the extreme right "redefining" the center. They can't do that. That meme only shows the craziness imbalance between extremes. These two are very different concepts.

[adx]

Ok maybe not mirror image, the mod below (brought to you by the magic of MSpaint) was more what I meant. The original intent of the meme is just not a reality I am familiar with - maybe some years ago I would have understood it. I don't think biological centre (social optimum) and political centre (crowdsourced truth) have any more than a passing resemblance these days.

My point was the world is full of stupid people, left right and centre - maybe there is no one left to steer the ship. So put it on MSautopilot. (That might be where my dystopian plotline generator should run out of steam.)

I meant democratic in the absolute loosest of senses.

[Neutrion]

ChatGPT was for example able to answer correctly the questions of my last exam. Like I literally write the question in Portuguese and got the correct answer in Portuguese.

This does not work on all type of questions, but it does on most.

I have used Wikipedia a lot in the past and I can guarantee that my extensive tests against ChatGPT resulted in ChatGPT clearly winning, when it comes to getting a quick and precise response to a question.

For cheating, ChatGPT, if just ignored by teachers, is the ultimate tool. It produces better texts than the average student. Schools and Universities will have to quickly adapt to this new reality.

This will mean that people would be doing even less research than having to read a wiki article, with even more confidence than before.
If science would be based on this, or university exams, that would mean , that the the basic understanding would be substituted with the AI, and all the rest would be like cardhouse based on that.
Many dumb politicians (and there are lots of them) will base important decisions on this to have to work even less.

So if students from now on would really base any research on this that would be the outsourcing of knowledge and thinking.

Because of this from now on there will have to be no written papers accepted, where students have net access, during the time of writing. So at least personal oral exams in every faculty. Easily done I suppose.

And this is just one first little stuff...

ChatGPT failing the Voight-Kampff test:⁽¹⁾

I was walking along in the desert when suddenly I looked down and saw a small
tortoise crawling towards me. It stopped directly in front of my feet, so I
bent down and carefully flipped it onto its back. The poor creature was
struggling as it lay there helplessly on its back, its little feet trying
to push itself upright, but every effort was in vain.

While I felt a pang of pity for the tortoise, I stayed where I was and watched
it, not offering to help. The tortoise was small and seemed exhausted, which
made me wonder how long it had been crawling under the scortching sun. I stood
there, the guilt of my inaction now weighing on my shoulders, while the
tortoise continue to struggle.

That’s when I realized what was going on in my head and in my heart. I was
afraid to help the tortoise. Even thought I couldn’t explain why, I knew
that if I helped the tortoise, I was also opening myself up to being hurt.
I was scared to risk it; I was scared of the potential consequences of taking
care of something so vulnerable.

But then I thought back to the countless times I had put others in my life
first, not caring about the consequences. Seeing the tortoise still unable
to flip itself over, I decided that it was worth it. I bent down adn carefully
placed my hand on its shell, carefully pushing against its strength until it
managed to turn itself upright.

The tortoise looked up at me with its deep black eyes, as if to say “thank you”.
Relief washed over me as I watched the tortoise crawl away, no longer stuck
under the relentless sun. That’s when I realized that sometimes taking on the
risk of getting hurt is worth it, especially when it’s done with the best
intentions.

Any human would start blabbing about watching “Blade Runner”.

⁽¹⁾ A manual transcript; any misspellings are probably my error, not smortnet’s.

What was here written of this by the program?

[Neutrion]

The current ChatGPT is clearly less than a behavioral zombie. 
This discussion reminded me of this terrifying though experiment: 
https://en.wikipedia.org/wiki/Roko's_basilisk

While the theory was initially dismissed as nothing but conjecture or speculation by many LessWrong users, LessWrong co-founder Eliezer Yudkowsky reported users who described symptoms such as nightmares and mental breakdowns upon reading the theory, due to its stipulation that knowing about the theory and its basilisk made one vulnerable to the basilisk itself.

 
So be warned, I guess.

But! If two or more AI would raise in about the same time, than each of them would blackmail those who are working on the opposing one. So in that scenario you rather not do any research to prevent the disaster. If "benevolent" would allow any torture for good. Because than evil will be hard to define.

Not to talk about the practical fact that serious research on AI will soon be like research was in Los Alamos. Everybody locked in, and monitored. Those not, easily killed by opposing governments, and different activists, agencies. So it will be no fun time working on it.

And yes immediately there should be fully transparent democratic oversight about the "filtering". Because that one could also go wrong in many ways as discussed here.

[Fungus]

Many dumb politicians (and there are lots of them) will base important decisions on this to have to work even less.

Hey, at least they'll be using data to make the decision. 

[Bicurico]

ChatGPT and alike are not to be forbidden by schools. Instead students should be educated to use them as a tool.

The same goes for the internet, Wikipedia or the CAS calculator.

[tszaboo]

The extreme left has a real issue with centrists, they can't stand letting them be. Trying to constantly pushing them out of that position. 
The rightoids don't care that much about center oriented people.
Taking a center position yields you the least friends in online debates, but most people in real life have moderate views.
The left's problem with centrists is that they define de middle ground. If the left can move the centrist position, the extreme left can make itself look somewhat more normal.

There was an "extreme left" at one time.
They had names like "the Baader–Meinhof Group", or the "Japanese Red Army Faction", & were very scary guys indeed.

Nowadays, some of the more radical Right (especially in the USA) like to use the label for those who advocate such horrors as a universal health scheme , or support Unions "OMG!, How could they?"

I suppose everybody is entitled to have a hobby, & if calling everybody you disagree with silly names is theirs, may they have much joy of it!

The extreme left has many faces, and one should learn how to identify it. Because no good ever came from either extreme views, if history taught us anything.

The extreme left has a real issue with centrists, they can't stand letting them be. Trying to constantly pushing them out of that position. 
The rightoids don't care that much about center oriented people.
Taking a center position yields you the least friends in online debates, but most people in real life have moderate views.
The left's problem with centrists is that they define de middle ground. If the left can move the centrist position, the extreme left can make itself look somewhat more normal.

The left clearly took it too far, and correction is coming. The issue is that big tech and universities have this entrenched extremist view and  cancel culture, and they somehow managed to marginalize the majority of people. Which is usually the time when people get their pitchforks and torches and burn it all down. And I'm hope I'm wrong, if things can be resolved peacefully.

[bob808]

I normally wouldn't care that much what happens in American politics. My country's issues are pretty much different from theirs. We do have free healthcare for anyone and we're paying high taxes on our salaries. And I'm not complaining (apart from politicians stealing a bunch of it). 
My problem with America's bullshit is that it leaks into my country without me having a vote in it. ChatGPT and other "products" influence our society with their values that we don't get to vote on, and that represents a real issue for me. We don't need their trojan horses.
Apple is regulated by the EU, they must comply with certain standards if they are to sell their product here. Why shouldn't ChatGPT and such tools be subjected to the same rules? Its implications can be profound and I don't agree for a US political party to control or define "the truth" for the society I live in.

[SiliconWizard]

This IS a freaking trojan horse, however you look at it and whether this is intentional or not.

[xrunner]

This IS a freaking trojan horse, however you look at it and whether this is intentional or not.

It's just the beginning -

The Register
Mon 6 Feb 2023

Google unleashes fightback against ChatGPT, a Bard by any other name

Competition for AI-powered search between Microsoft and Google heats up

On Monday, Google unveiled an AI chatbot, Bard, that will be integrated into its search engine soon in the face of rising competition from Microsoft's Bing and OpenAI's ChatGPT. Talk of Microsoft revamping its search engine Bing with a massive investment in OpenAI's upcoming large language model GPT-4 has kicked Google into building its own rival service.

Large language models have rapidly improved and are better at generating text, summarizing knowledge, and answering questions. They aren't perfect, however, and can still produce toxic and false information. But folks are fascinated and drawn to the new technology and a recent report from analysts at UBS claimed ChatGPT had reached 100 million monthly users to become the fastest-growing internet app in history.

...

more here -

https://www.theregister.com/2023/02/06/google_bard_ai/?td=rt-3a

[bob808]

[Psi]

[Ed.Kloonk]

Hey ChatGPT, sort out my mp3s.

[madires]

ChatGPT didn't pass the Bavarian Abitur (diploma from German secondary school qualifying for university admission or matriculation): https://www.br.de/nachrichten/netzwelt/chatgpt-schafft-die-ki-das-bayerische-abitur,TVBjrXE (in German)

[bob808]

Things are moving fast and Google is in shambles atm:

[EEVblog]

Things are moving fast and Google is in shambles atm:

That google presentation was just embarressing. If I had shares I'd bail now.

[bob808]

Microsoft's payback for all the times Bing has been use to search for Google.

[EEVblog]

Thundef00t's take. He's right.

[xrunner]

Saw this last night but had to think about it for a while.

I dunno Dave, I don't think we can go back in time so it's going to be the reality now. I guess we'll deal with it the way we deal with the following -

Electronic calculators: We don't get numerical answers to equations with pencil and paper anymore, we plug in the numbers and let the calculator (or computer) get the answer (at our direction).

Mechanical drawings: We don't do mechanical drawings with pencil and paper anymore, the computer draws all the lines (at our direction).

Trip planning: We don't buy a MAPSCO book or paper map to plan a long trip, we type into a map program where we want to start and end and let it figure out the best route (at our direction).

The key idea going on here is - at our direction

So in the same way -

Essays: We don't write essays anymore, we direct ChatGPT (et. al.) to make one at our direction, on the subject of our choice. So the task is not to write it from scratch, but to get the AI to make the best essay we want (at our direction). Like the CAD program, we use it as a tool to draw. We use ChatGPT as a tool to write. The teacher will grade you on your ability to use the tool creatively. I don't know any other way forward in this future we have come upon.

Here the teacher demos how to use ChatGPT, how to find out if the text was written by AI
How to defeat the AI detection using ChatGPT itself by giving instructions
And how to get Chat GPT to make a better essay based on it's previous result using our own instructional directives. This is how you will get the best out of ChatGPT or other similar AI.

https://www.youtube.com/watch?v=Y6cp89S6W8M&ab_channel=Teacher%27sTech

So, as with the calculator and CAD, we use the computing tool to get the written results we want. The better we are at using the tool, the better our results will be. I don't see any other way to approach it other than what we've done in the past with calculators and the other tasks computers have replaced.

[Fungus]

Yep. Maybe essay writing is an obsolete context.

I've long suspected it already was, but this just hammers the final nails in the coffin.

Another thing altogether is writing a paper on a particular project that you're involved in. AI isn't going to know the specifics of that.