I had a conversation with deepseek AI about pH

[stdsch40]

So I had a conversation with deepseek-r1:1.5b about pH. I have it running on my single board computer (not really up to the task but whatever) and asked it a few simple questions. This is what it came up with...

 

[stdsch40]

My son asked deepseek AI about black carpenter ants...it said that insects have 8 legs. We looked at each other and thought wtf.

 

[MarkB_MI]

How well does autocorrect work on your phone? It's probably right about half the time, and way off base the rest of the time. That's because it's just guessing about the word you want, based on what it thinks the probability is of that particular word coming up next. Today's so-called artificial intelligence systems are basically autocorrect on steroids. The AI systems don't 'think' in the way that humans or even most animals think. AI systems vacuum up a huge amount of information, and from that stored information try to predict the 'right' answer to your question. As your little test demonstrated, they're right part of the time, but when they're wrong they're really wrong.

That said, AI today is really in its infancy. It's fairly frightening to think where it might be in a few years. In the meantime, we're going to have some serious consequences from folks relying on AI when they shouldn't. I'm finding a lot of stuff on the web these days that appears to have been authored by either AI or idiots. And, given that AI is cheaper to use than idiots, it's probably the former.

 

[showcrop]

I would think that AI would use the correct "try to" in stead of the colloquial "try and" which of course doesn't actually mean what you want to say.

 

[joeysmith]

AI is akin to GPS
You better use a little common sense with it or you'll be lost

 

[DRussell]

So I had a conversation with deepseek-r1:1.5b about pH. I have it running on my single board computer (not really up to the task but whatever) and asked it a few simple questions. This is what it came up with...View attachment 102194

You should say next, "If you think that is a strong correction, then you don't understand the meaning of correction.

 

[Dober_paw]

Wo

AI is akin to GPS
You better use a little common sense with it or you'll be lost

Agreed - but not just simply lost, it's more like infected and now with the potential to spread disinformation.
I wonder how it would respond if it was in a position requiring it to "lose face" or to question authority? Oh, wait, that would suggest it was sentient not just a piece of software and statistics.

 

[dcarp]

So I had a conversation with deepseek-r1:1.5b about pH. I have it running on my single board computer (not really up to the task but whatever) and asked it a few simple questions. This is what it came up with...View attachment 102194

Being that deepseek came from the land of almost right its answer seems right inline.
It would seem that they “acquired” the technology to produce this AI and cut as many corners as possible to rush it out to show the US up.

 

[Watthour]

My son asked deepseek AI about black carpenter ants...it said that insects have 8 legs. We looked at each other and thought wtf.

It may have taken several computational cycles just to filter through the synonyms, antonyms, relative and comparative terms and phrases, colloquialisms, and definitions before determining that your inquiry may have been the polar opposite of black carpenter uncles. That begs the questions about the distinctions between that and asian carpenter relationships, latin carpenter relatives, caucasian carpenter kin, and other ethnicities swinging hammers. No wonder AI is so befuddled.

 

[Watthour]

How well does autocorrect work on your phone? It's probably right about half the time, and way off base the rest of the time. That's because it's just guessing about the word you want, based on what it thinks the probability is of that particular word coming up next. Today's so-called artificial intelligence systems are basically autocorrect on steroids. The AI systems don't 'think' in the way that humans or even most animals think. AI systems vacuum up a huge amount of information, and from that stored information try to predict the 'right' answer to your question. As your little test demonstrated, they're right part of the time, but when they're wrong they're really wrong.

That said, AI today is really in its infancy. It's fairly frightening to think where it might be in a few years. In the meantime, we're going to have some serious consequences from folks relying on AI when they shouldn't. I'm finding a lot of stuff on the web these days that appears to have been authored by either AI or idiots. And, given that AI is cheaper to use than idiots, it's probably the former.

A responsible adult person near you has noticed a similar phenomenon:

 

[Phill Tibbetts]

Who really gives a rats ars.

 

[stdsch40]

Who really gives a rats ars.

Want to know why you should 'give a rat's ars'? Because the corporations and governments of this world are going to inflict this shyt upon all of us unless we do what we can to stop it. Educate yourself and your children without AI, don't believe everything the hucksters and grifters tell you. 'be wise like snakes and innocent as doves'.

 

[will-max dairy]

AI is akin to GPS
You better use a little common sense with it or you'll be lost

If AI is completely akin to GPS, we have to think about GPS's feature of "selective availability"... there was a time when the accuracy of the GPS signal could be throttled by the introduction of intentional errors, and only recievers with a "crypto key" were privy to an alternate stream of information that gave the intentional errors, so that they could be subtracted out.

This was preserved to give a competitive advantage to our military in times of conflict.

AI could very easily have the same type of accuracy throttling or "curation," if you want to call it that... implemented...so that some people are always just a couple of steps behind.

When the business credo is "buy on the idea, sell on the news..." metering the rate that people have access to ideas would be a very viable method of economic control.

It makes me think about online gambling. There are "apps" that you can buy that supposedly use AI to help you place bets on everything from horseracing, to football to how long you'll wait for your coffee at Starbucks.

If you think that the companies setting the lines on those bets are going to sell the same level of prediction that they use... I have a bridge that I wanna sell you.

PS. The notion of "selective availability" of GPS is a very 90's concept... back when I was working on it for the AF. Thirty years ago, to even mention that was classified. Thirty years ago, placing a GPS receiver on a surveyed spot and broadcasting the error, even locally, was a criminal act. Supposedly, GPS has become so ubiquitous now, that we don't even rely on selective availability...

 

[stdsch40]

“Behold, I send you out as sheep in the midst of wolves. Therefore be wise as serpents and harmless as doves."

 

[MarkB_MI]

Want to know why you should 'give a rat's ars'? Because the corporations and governments of this world are going to inflict this shyt upon all of us unless we do what we can to stop it. Educate yourself and your children without AI, don't believe everything the hucksters and grifters tell you. 'be wise like snakes and innocent as doves'.

Because the corporations and governments of this world are going to inflict this shyt upon all of us unless we do what we can to stop it.

Hmm. What exactly do you propose to do to 'stop' AI?

 

[BarnyardEngineering]

 

[GChief]

View attachment 102240

Hey, at least they keep my REYN stock going

 

[will-max dairy]

Hmm. What exactly do you propose to do to 'stop' AI?

With things like this, it's more about setting boundaries on your own behavior. There is no way that we can "stop" AI; but we can pick and choose how we participate with it or in it.

Like the comment above about the accuracy of autocorrect. More recent implementations of excel try to recognize patterns as you're creating spreadsheets to autofill subsequent cells. It's about 50/50, in my experience. I've even had to correct a report done by one of our interns; because excel automatically summed cells in a column that shouldn't have been summed. She didn't want to sum them. She knew enough not to; but I went over and sat with her for a moment, and the version of excel provided by MS Teams had a feature that it would perform an autosum if you clicked and dragged along a column to, for instance, do a paste operation. No bueno!

When we routinely create technical reports with thousands of spreadsheet cells that should be accurate, the notion that a systemic errors could be randomly sprinkled throughout a worksheet or a workbook is something that we need to avoid.

On the other hand, using spell check and grammar check... and picking and choosing the corrections to accept or reject, is very handy.

So...my boundary here is... I turn off autocorrect and only use it when I want it. I turn off many of these automatic features in excel and only use them when I want them.

I also never think that AI will ultimately give me an advantage with things like gambling or the stock market. The entities that control such things have access to more information and better algorithms than I, so I tend to avoid participating in such curated fields.

 

[MarkB_MI]

With things like this, it's more about setting boundaries on your own behavior. There is no way that we can "stop" AI; but we can pick and choose how we participate with it or in it.

Like the comment above about the accuracy of autocorrect. More recent implementations of excel try to recognize patterns as you're creating spreadsheets to autofill subsequent cells. It's about 50/50, in my experience. I've even had to correct a report done by one of our interns; because excel automatically summed cells in a column that shouldn't have been summed. She didn't want to sum them. She knew enough not to; but I went over and sat with her for a moment, and the version of excel provided by MS Teams had a feature that it would perform an autosum if you clicked and dragged along a column to, for instance, do a paste operation. No bueno!

So...my boundary here is... I turn off autocorrect and only use it when I want it. I turn off many of these automatic features in excel and only use them when I want them.

I also never think that AI will ultimately give me an advantage with things like gambling or the stock market. The entities that control such things have access to more information and better algorithms than I, so I tend to avoid participating in such curated fields.

With things like this, it's more about setting boundaries on your own behavior. There is no way that we can "stop" AI; but we can pick and choose how we participate with it or in it.

Yes, we can choose not to use AI. But we can't stop others from using it, even though we may be profoundly affected by their use of AI. One little AI glitch could stop us from getting a job, or qualifying for a loan.

've even had to correct a report done by one of our interns; because excel automatically summed cells in a column that shouldn't have been summed.

In my experience, most of the folks who think they know how to use Excel don't know how to use Excel. I don't expect these same folks to be any better when they start using AI.

I also never think that AI will ultimately give me an advantage with things like gambling or the stock market. The entities that control such things have access to more information and better algorithms than I, so I tend to avoid participating in such curated fields.

Agreed. Never bet against the house.

 

[will-max dairy]

MarkB_MI wrote:

"Yes, we can choose not to use AI. But we can't stop others from using it, even though we may be profoundly affected by their use of AI. One little AI glitch could stop us from getting a job, or qualifying for a loan."

Agree. A "boundary" is about our behavior, and what we personally will do. It the situations above, my "boundary" is to simply know that AI is now becoming the invisible hand behind many diagnostic, screening and qualification processes. This means that I consider anything about my situation that would be an outlier that wouldn't fit into a set algorithm for these processes. If I can't think of an outlier, then the result passes the "smell test" and I just accept the result.

AI is also ripe to hand many people degrees from Dunning and ------- university. (for some reason... the word K*r*u*g*e*r seems to be turned into dashes after I submit my post... I have no idea why.)

Myself, I think it's kinda fun. Interacting with AI is much like having a child around to ask questions and bounce ideas off of. Sometimes, you'll get a gem. Most times you don't. Many times you'll get a chuckle.