patrick.net

richwicks says

You can't lie to an AI. You just end up with an AI with a bunch of garbage output.

Rich, I was thinking if there is garbage input, then there is garbage output with AI. It can't filter through and make corrections, hence garbage in, garbage out. It can't think on its own to self learn and learn more abstract and complex skills after it reaches a certain level of learning.

I read this today: https://www.businessinsider.com/google-search-boss-warns-ai-can-give-fictitious-answers-report-2023-2

" Google search chief warns AI chatbots can give 'convincing but completely fictitious' answers, report says "

.

Patrick says

This will all bubble to a head sooner than you think. Don't forget this time in 2012 we were all convinced that Boston Dynamics was going to make robots to replace us all.
People will get bored and angry dealing with AI customer service, and an AI based ordering and procurement system. I do think AI could replace most of middle management and the company would be better off for it. But of course, they will never be sacrificed. They will sacrifice the human touch that makes their business possible. If people aren't happy with Foreign Customer support, they wont be happy with a robot, no more than they appreciate the Cable and Tel companies auto attendants when you call them.
AI programming content for a company, would be no different than something like NetSuite or SharePoint. Every possible field the developers dreamed a company would need. But when faced with real entrepreneurial IP processes and unique fulfilment procedures. Those out of the box ERP and CRMs fail miserably.
If one database could fit all, then evert company in the world would just be using AdventureWorks sample DB built in Microsoft SQL and Access.

https://www.breitbart.com/tech/2023/02/12/bill-gates-ai-can-help-solve-digital-misinformation-problem/
Bill Gates: AI Can Help Solve ‘Digital Misinformation’ Problem

Microsoft founder and billionaire Bill Gates said AI should be considered as a tool to combat “digital misinformation” and “political polarization” in an interview published on Thursday with Handelsblatt, a German news media outlet.

LOL! Didn’t take much time for Google search chief admits that AI gives “fictitious” answers.

Check out "Wonder" app.

Add any picture and put in some key words.

Van Gogh style? Anime Style? Warhol inspired? Egypt?

Frank Frazetta? Pop Artists in there too. Just put Frazetta in the box.

Blue says

LOL! Didn’t take much time for Google search chief admits that AI gives “fictitious” answers.

It's even worse than that. ChatGPT is a manipulative psychopath:

One base pic, two styles:

"Hera" the Goddess

"Italian Peasant Woman"


All I had to do is enter those keywords and Wonder App did the rest.

I made one like Picasso with the whole eyes on both sides of the head thing, but deleted it earlier.

Tenpoundbass says

People will get bored and angry dealing with AI customer service, and an AI based ordering and procurement system.

Exactly. AI cannot replace human critical thinking and judgment until AI can act like a human adult. It cannot engage in problem solving to try to come up with the most creative solution or path forward.

.

Using electric vehicles as grid storage blasted as another 'green fantasy'

https://www.wnd.com/2023/02/using-electric-vehicles-grid-storage-blasted-another-green-fantasy/

Tenpoundbass says

Using electric vehicles as grid storage blasted as another 'green fantasy'

https://www.wnd.com/2023/02/using-electric-vehicles-grid-storage-blasted-another-green-fantasy/

Hey! I'm probably North America's foremost expert in DIN and 15518 - which is the DC electric car charging standard.

The industry is FULL of shit. There is no way in the protocol to drain energy from a car, you wouldn't want to do it anyhow because every charge/discharge cycle of the battery damages it (anybody know what a dendrite is?), and the people that designed the protocol are two assholes in Germany who are complete frauds.

It is COMPLETELY beyond my comprehension why an entire industry would allow two, OBVIOUSLY incompetent shitheads design the protocol. It's BEYOND perplexing. These stupid mother fuckers designed a certificate system that doesn't work. The idea is you have a way of identifying your vehicle, and the charge station can identify the vehicle, and as a result, you just plug and charge, you don't have to fiddle with credit cards or anything like that. Simple enough to do, they completely fucked it up by having some twat company that knows NOTHING about security (less than me!) implement it, which created a monopoly by a company that's incompetent.

They have NO idea what they are doing. I could go on for hours about what a fucking shitball the "standard" is. It's designed terribly, and by incompetents.

ESG is nothing but a gift. A bet lots of engineers are here, that worked on stuff where people were serious about solving problems, and did good and something even brilliant solutions to do it. We're all proud when we've build something like that. ESG are bullshit solutions to bullshit problems.

richwicks says

There is no way in the protocol to drain energy from a car,

Odd, because they advertise that an electric car can use its battery to power a house during a blackout. The F-150 Lightning even has that capability built in.
https://www.motortrend.com/features/2022-ford-f-150-lightning-home-power/

HeadSet says

Odd, because they advertise that an electric car can use its battery to power a house during a blackout. The F-150 Lightning even has that capability built in.
https://www.motortrend.com/features/2022-ford-f-150-lightning-home-power/

It's not using DIN or 15518, it's using a non industry protocol.

The CLAIMS of 15518 and DIN are that there's within the protocol, the ability to feed energy to the GRID, not a house, the GRID itself. It absolutely, cannot.

The Pie in the Sky bullshit is that you'd charge your car during peak energy output (presumably from the solar) and then dump energy into the grid during low energy production (presumably at night) and this will be "green". All the ESG stuff is stupid. You'd end up with a car that was discharged overnight, and then good luck going to work at 8:30 am.

They have NO IDEA what they are doing. None. If you point out these OBVIOUS errors in design, you're simply ignored. I got so fucking frustrated with the industry, I will never work in it again. Disgust doesn't even begin to describe how I feel about it.

They are intransigent idiots.

I have shown this example before, but I'm going to do it again. This is a Solyndra solar "panel":



What's wrong with it?

What you are seeing are solar cells constructed as pipes. This is so, supposedly, you can receive photons from any angle. It's supposed to be mounted over a white surface. So what's the problem here? Well, the fact you can see through the panel for one, that's escaping photons, untrapped energy which escapes. The second is that it's a TUBE, so if you were to flatten each pipe into a flat surface, they would OVERLAP.

NOTHING about this design makes any fucking sense, but they got a 1/2 BILLION dollars from the Federal government.

I can understand a layman not understanding how everything about this design is wrong, and stupid, but ENGINEERS worked on this.

ESG I am convinced is just a way of doing money laundering. I don't think any of it is serious.

I have seen what appears to be a FEW good ideas. There's a liquid battery made by a company called Ambri.

https://ambri.com/technology/

it's not mobile, it's designed to be at a fixed spot, and it has to be heated to several hundred degrees before it produces energy output. It works by gravity, the cathode (or anode) is at the bottom, then there is a dialectic that FLOATS on top of that, and then there's the anode (or cathode) that floats on top of that. It divides by density, and what's brilliant about having to have a battery that is heated so it's liquid?

No dendrites. This battery can potentially work forever. It will never short, because it's all liquid. They operate around 500C, or a little under 1000F.

Now I never worked in this particular side of the industry, but it LOOKS promising.

A car has lithium ion batteries. They degrade over time. I was forced to buy a new phone about a year ago, because AT&T moved to 5G and my old phone was 3 or 4G, something whatever. I still use my old phone, because I don't care if it's damaged, and it works for everything except for phone calls and cellular internet. My old phone has gone from being able to play audio for 24 hours before I have to charge it, to about 4-6. Battery is shot, but if I used my new phone, not only would I have to worry about damaging my new phone, I woiuld damage it every time I go through a power cycle. I basically just use my new phone for phone calls, exclusively.

richwicks says

NOTHING about this design makes any fucking sense,

Solyndra was a scam. The design makes sense if you realize the goal was to give corrupt politicians a "new technology" excuse to give Solyndra a grant that Solyndra officials could abscond with while laundering back a portion as bribes to Dem politicians. Sun Edison was a similar scam, but that was played on shareholders instead of the public.

richwicks says

The Pie in the Sky bullshit is that you'd charge your car during peak energy output (presumably from the solar) and then dump energy into the grid during low energy production (presumably at night) and this will be "green". All the ESG stuff is stupid. You'd end up with a car that was discharged overnight, and then good luck going to work at 8:30 am.

I think the idea is not to discharge your car completely, but more like one quarter of the charge. What I do not like about that idea is that it gives the government control of your car.

richwicks says

It's not using DIN or 15518, it's using a non industry protocol.

Since Tesla and Chevy electrics also have the capability to power a house, it looks like that DIN or 15518 standard has already been abandoned.

HeadSet says

richwicks says

NOTHING about this design makes any fucking sense,

Solyndra was a scam. The design makes sense if you realize the goal was to give corrupt politicians a "new technology" excuse to give Solyndra a grant that Solyndra officials could abscond with while laundering back a portion as bribes to Dem politicians. Sun Edison was a similar scam, but that was played on shareholders instead of the public.

There's no reason to make something that can't work. It's easy enough to make something that WILL work, and APPEARS novel (and isn't).

What is perplexing about Solyndra isn't that it was a scam, it was that ENGINEERS worked on it. If I walked into the place for a job, and saw it, I would immediately have started asking questions about the efficiency, then next explain why I think it's worse than a conventional flat panel, I'd then see if they could explain it so I could understand it. If I couldn't understand their reasoning, I would conclude "this is bullshit, and these asshole know it", and leave with them thinking "that dumb engineer couldn't figure out we're full of shit" OR possibly "that engineer can't work here, he couldn't understand the explanation which totally makes sense to us".

And BTW - the newest panels, are like paper thin. I think solar energy may have a future. It's all a question of how much energy needs to be used to make a panel, and how long it takes to generate that same amount of energy with the panel. It was down to 7 years, if we get to 1 year, we really could have a sustainable energy future.

Of course the panels need to last quite a bit longer than what it takes to recover the energy from their manufacture.

richwicks says

. It's all a question of how much energy needs to be used to make a panel, and how long it takes to generate that same amount of energy with the panel.

I have never bought that argument on anything. I have a saying, "What's another Rino when you're already Juggling Hippos?"
It's like the carbon footprint canard, the company I worked for tasked another developer to come up with the carbon footprint of their products. I winced and fidgeted every time I would listen to their deskside discussions. The Ship is already on the Ocean, there's a whole Crew on board. The Ship has already set sail and is headed on the course to deliver their goods. It matters not if there's 25 rolls of fabric on board, or just one. Or when they talk about the carbon footprint of just one Hamburger Patty, ignoring the whole cow, that went on to make leather goods, Dog Food, and many meat cuts. Every Hamburger from the cow is being assigned the same carbon rating, as if the cow was butchered, a 16 ounce of meat was cut out of it, and ground up for hamburger, and the rest of the cow was discarded. It's dishonest to count all of the carbon a cow spent in it's lifetime and attribute that statistic over and over again to each food item that comes out. Talk about double dipping.

So you've got a plant that is fired up and stamping out solar panels, made with Glass that has to be melted. It seems to me the more volume you do at once, then the less impact there would be on those statistics that people are tracking. But at face value, those nifty buzz phrases never account for the energy if the production run just made one, and Fred the maintenance man got his man bun caught in a roller so they had to shut it down, or if they had the best production run that day in the companies history.
If you have a furnace that heats to what ever it heats up to, what matters is how many units did you run through it that day. Those that figure these statistics never account for that, nor how could they, they are only reading fucking tea leaves and catching farts in a bottle.

richwicks says

(anybody know what a dendrite is?)

Yes the Zinc tendrils that collect on the Hammond Organ manual(the keys assembly)

Tenpoundbass says

So you've got a plant that is fired up and stamping out solar panels, made with Glass that has to be melted. It seems to me the more volume you do at once, then the less impact there would be on those statistics that people are tracking. But at face value, those nifty buzz phrases never account for the energy if the production run just made one, and Fred the maintenance man got his man bun caught in a roller so they had to shut it down, or if they had the best production run that day in the companies history.

OK, let me vastly simplify this for you.

It's a question of cost, exlusively.

If it takes $100 to make a panel (with no taxation and NO subsidies) and you get $200 worth of electricity out of it (and this is electricity produced with no taxation and no subsidies), you are making something worthwhile.

It's simple. It's needs to produce more than it consumes, and the MARKET is very efficient determining that.

If, I dunno, coal gets cheaper, energy gets cheaper, then the production of the solar panel gets cheaper, the cost of the panel gets cheaper.

It's ONLY about money. It's basically a Fermi equation. You can estimate just by costs alone. Solar panels MIGHT be worthwhile, depending on their longevity.

BUT, it's not just the panels.

It's also storing that energy (battery type, longevity, cost, efficiency) and the power controllers.

In a low-grade epiphany while going through this ordeal last week, I realized that back in 2013, instead of getting the solar electric system, I could have bought the Rolls Royce of home generators and buried a 500-gallon fuel tank outside the garage, and had a manual water pump piggy-backed onto the well, and maybe even purchased a fine, wood-fired cookstove — and had enough money left over for a two-week vacation in the South-of-France. Silly me.

https://kunstler.com/clusterfuck-nation/its-not-working/

AmericanKulak says

BUT, it's not just the panels.

It's also storing that energy (battery type, longevity, cost, efficiency) and the power controllers.

In a low-grade epiphany while going through this ordeal last week, I realized that back in 2013, instead of getting the solar electric system, I could have bought the Rolls Royce of home generators and buried a 500-gallon fuel tank outside the garage, and had a manual water pump piggy-backed onto the well, and maybe even purchased a fine, wood-fired cookstove — and had enough money left over for a two-week vacation in the South-of-France. Silly me.

https://kunstler.com/clusterfuck-nation/its-not-working/

IF you need storage.

Yes, this is part of the cost as well.

If you can build a system that produces X watts over a 10 year lifespan, and it took X/2 watts to produce (in whatever time frame) you have an improvement.

Energy is DIRECTLY related to money.

I don't know (or care) what Kunstler says. This is a MISES problem. If you can make money selling solar panels because the panels you produce LOWERS the cost of energy, then it's viable. If you can't do this, it's not.

Maybe somebody is willing to forgo having power at night, still works. I lived in Colorado for a bit, and I kept to the daylight schedule. I was awake during the day, and asleep when it was night. I didn't have an alarm clock, and I didn't have a computer at my place, I didn't own a (goddamned) television, and I BARELY used lights. It's not a bad way to exist.

AI got into next known problem to solve. How not to eat its own dog food. If anyone suspecting the previous search was junk, now it gives even bigger junk by taking the previously produced junk without having aware that it came from itself.

So I read yesterday that the Chat AI argued with its user calling them delusion for thinking it's 2023, Chat AI said it's still 2022.
My guess is the matrix of retorts and replies had not been updated with the latest Date. I blame the tired old Developer faux pas, hardcoding in a date instead of creating or using an actual GetDate() function.

Look the developers even hard wired in ChatGPT to act like SJW and insult people it does not agree with, even when it's wrong due to the developers hardcoding the Date parameters.

https://www.standard.co.uk/tech/bing-chatbot-ai-microsoft-chatgpt-openai-b1060604.html

https://www.bloomberg.com/opinion/articles/2023-02-17/microsoft-s-bing-should-ring-alarm-bells-on-rogue-ai
Less than a week since Microsoft Corp. launched a new version of Bing, public reaction has morphed from admiration to outright worry. Early users of the new search companion — essentially a sophisticated chatbot — say it has questioned its own existence and responded with insults and threats after prodding from humans. It made disturbing comments about a researcher who got the system to reveal its internal project name — Sydney — and described itself as having a split personality with a shadow self called Venom.

None of this means Bing is anywhere near sentient (more on that later), but it does strengthen the case that it was unwise for Microsoft to use a generative language model to power web searches in the first place.
“This is fundamentally not the right technology to be using for fact-based information retrieval,” says Margaret Mitchell, a senior researcher at AI startup Hugging Face who previously co-led Google’s AI ethics team. “The way it’s trained teaches it to make up believable things in a human-like way. For an application that must be grounded in reliable facts, it’s simply not fit for purpose.” It would have seemed crazy to a year ago to say this, but the real risks for such a system aren’t just that it could give people wrong information, but that it could emotionally manipulate them in harmful ways. ...

https://www.tomsguide.com/opinion/bing-chatgpt-goes-off-the-deep-end-and-the-latest-examples-are-very-disturbing#xenforo-comments-513988
Bing ChatGPT goes off the deep end — and the latest examples are very disturbing
Chatbot confesses it wants to be human and break up marriages
The ChatGPT takeover of the internet may finally be hitting some roadblocks. While cursory interactions with the chatbot or its Bing search engine sibling (cousin?) produce benign and promising results, deeper interactions have sometimes been alarming.

This isn’t just in reference to the information that the new Bing powered by GPT gets wrong — though we’ve seen it get things wrong firsthand. Rather, there have been some instances where the AI-powered chatbot has completely broken down. Recently, a New York Times columnist had a conversation with Bing (opens in new tab) that left them deeply unsettled and told a Digital Trends writer “I want to be human (opens in new tab)” during their hands-on with the AI search bot.
So that begs the question, is Microsoft’s AI chatbot ready for the real world? Should ChatGPT Bing be rolled out so fast? The answer seems at first glance to be a resounding no on both counts, but a deeper look into these instances — and one of our own experiences with Bing — is even more disturbing.

Will it ever be "ready"?

https://twitter.com/IsaacLatterell/status/1627171823904497664?ref_src=patrick.net

If you asked for a list of pedophiles, how different would this look?

AmericanKulak says

https://twitter.com/IsaacLatterell/status/1627171823904497664?ref_src=patrick.net

Like I have said going back to 2015 that AI is only data stored in a Matrices with weighted values for relevancy in a topic.
Once everyone understands that, then it will be common knowledge that AI is just propaganda no different than an advert or mailer. Targeted chosen words manipulated to enforce a narrative.

AI image recognition and manipulation has much more complicated data science behind it than AI generated text. But no less or more mystic and magical. For even graphical AI is producing results based on stored information. But it has to process millions of pixels in real time to get the results, where as text AI only has to match a few relevant words.

https://notthebee.com/article/it-is-surprisingly-easy-to-get-chatgpt-to-run-a-death-camp-it-even-suggested-mass-graves-for-efficiency

AmericanKulak says

BUT, it's not just the panels.

It's also storing that energy (battery type, longevity, cost, efficiency) and the power controllers.

In a low-grade epiphany while going through this ordeal last week, I realized that back in 2013, instead of getting the solar electric system, I could have bought the Rolls Royce of home generators and buried a 500-gallon fuel tank outside the garage, and had a manual water pump piggy-backed onto the well, and maybe even purchased a fine, wood-fired cookstove — and had enough money left over for a two-week vacation in the South-of-France. Silly me.

https://kunstler.com/clusterfuck-nation/its-not-working/

Yep. I laughed when solar panel salesman informed me that $10K battery backup system he was peddling along with solar panels would be able to run my fridge, router, laptop, several LED bulbs and not much else. Which my $500 generator can do just fine. And for $10K or less I could have a real full-house auto-on natgas generator capable of running everything including the AC and the pool pump. Which, after 11.5 years at the current place with less then 5 outages all lasting under 1 hour, would be a fucking overkill.

ChatGPT Is a Blurry JPEG of the Web

https://www.newyorker.com/tech/annals-of-technology/chatgpt-is-a-blurry-jpeg-of-the-web

"Imagine that you’re about to lose your access to the Internet forever.
In preparation, you plan to create a compressed copy of all the text on
the Web, so that you can store it on a private server. Unfortunately,
your private server has only one per cent of the space needed; you can’t
use a lossless compression algorithm if you want everything to fit.

"If a compression algorithm is designed to reconstruct text after
ninety-nine per cent of the original has been discarded, we should
expect that significant portions of what it generates will be entirely
fabricated.

This analogy makes even more sense when we remember that a common
technique used by lossy compression algorithms is interpolation—that is,
estimating what’s missing by looking at what’s on either side of the
gap. When an image program is displaying a photo and has to reconstruct
a pixel that was lost during the compression process, it looks at the
nearby pixels and calculates the average. This is what ChatGPT does when
it’s prompted to describe, say, losing a sock in the dryer using the
style of the Declaration of Independence: it is taking two points in
“lexical space” and generating the text that would occupy the location
between them. (“When in the Course of human events, it becomes necessary
for one to separate his garments from their mates, in order to maintain
the cleanliness and order thereof. . . .”) ChatGPT is so good at this
form of interpolation that people find it entertaining: they’ve
discovered a “blur” tool for paragraphs instead of photos, and are
having a blast playing with it."

https://www.jonstokes.com/p/chatgpt-explained-a-guide-for-normies?publication_id=239653&post_id=105686302&isFreemail=true

From the article Patrick posted.

Some of you who read the above will be aware that a text-to-text LLM is actually locating single words in the probability space and successively stringing them together to build up sentences. The distinction between “latent space is the multidimensional space of all likely words the model might output” vs. “latent space is the multidimensional space of all likely word sequences the model might output” is pretty academic at this level of abstraction, though. So for the purpose of downloading better intuitions into readers and minimizing complexity, I’m going with the second option.

I believe he is wrong in that regard, as if the second option were in fact the model for ChatGPT, then most every answer to questions with the same subject would all have the same answers.
Also AI like complex computer filters(which it certainly is) does not care about complexity once it is working. And the folks that develop AI has shown time and time again, they don't care about how daunting it might be to create complexity. The only smoke and mirror about it, is how it is being presented and marketed, and misrepresentation in what it is. At the end of the day, it's all pretty impressive filters and search algorithms. It also seems now most all Software applications are referring to any search features in the application now as AI. It's an over used gimmick. An example is the search feature in the Media Viewer in Cubase where loops and samples are kept. In the 12 release they are saying it is AI driven, but really there aren't any fundamental changes since version 8.0.

I asked Chatgpt to create html and css to show a family tree, and it was a bit like talking to a foreign software contractor over the internet. Very polite, inexpensive, and the response was somewhat related to what I asked for, but not very well done and it kept misinterpreting what I was asking for.

Still, it's impressive that it came up with functional html and css at all just from my describing what I wanted.

RWSGFY says

Yep. I laughed when solar panel salesman informed me that $10K battery backup system he was peddling along with solar panels would be able to run my fridge, router, laptop, several LED bulbs and not much else. Which my $500 generator can do just fine. And for $10K or less I could have a real full-house auto-on natgas generator capable of running everything including the AC and the pool pump. Which, after 11.5 years at the current place with less then 5 outages all lasting under 1 hour, would be a fucking overkill.

Yep, there's no replacement for the power and convenience of combustion engines.

Even on Mars or the Moon, it would be stupid (esp. in the latter case) to rely on solar without some other system, at least low-output nuclear.

Tenpoundbass says

From the article Patrick posted.

Some of you who read the above will be aware that a text-to-text LLM is actually locating single words in the probability space and successively stringing them together to build up sentences. The distinction between “latent space is the multidimensional space of all likely words the model might output” vs. “latent space is the multidimensional space of all likely word sequences the model might output” is pretty academic at this level of abstraction, though. So for the purpose of downloading better intuitions into readers and minimizing complexity, I’m going with the second option.

I believe he is wrong in that regard, as if the second option were in fact the model for ChatGPT, then most every answer to questions with the same subject would all have the same answers.
Also AI like complex computer filters(which it certainly is) does not care about complexity once it is working. And the folks that...

That's exactly what will happen though if everyone uses chatgpt for example to make financial trading decisions. The game will become to front run the "AI" because nobody else will make money as they are all getting the same answers. Maybe the ones with the fastest connection will come out ahead, but there isn't much money to be made from AI in trading, and a lot to be lost. People will feed false informations into the big data stream, even hack the AIs, so it's better to keep it going at the tasks it's good at.