What does a TRUMP presidency look like?

[crazy canuck]

Most of the risks around AI would go away if marketing people stopped pretending it actually thinks.

Yes, but to make the statement more accurate I would remove the word marketing and replace "pretending" with "believing". 

[Tamas]

Most of the risks around AI would go away if marketing people stopped pretending it actually thinks.

Yes, but to make the statement more accurate I would remove the word marketing and replace "pretending" with "believing". 

Good point.

[Josquius]

I'm torn on AI overall.
I used to be very much of the opinion it was all marketing buzz and would soon blow over.
Increasingly though I am worried. Its not in that place yet. I do still expect in the next year or two a backtracking on all those replace jobs with AI moves...
But longer term when it comes down to it.... Sure its not actually thinking and its just 1s and 0s throwing out the most likely stuff...but isn't that what humans are too?
The other day I had quite the existential panic over the future.

[crazy canuck]

No, neither generative AI nor humans throw out what is most likely. Generative AI fabricates things and what it generates has no relationship to anything approaching a probability analysis.  People also, with rare exceptions of those of us who are trained to think in terms of probabilities (like DGuller), do not think about what is most probable. People generally try to make conclusions about what is the best or correct course of action. Something that generative AI programs are also incapable of doing.

[viper37]

Trump Treasury Secretary Reveals Humiliating Detail About Tariff Talks

Bessent made the startling admission Sunday as he tied himself in knots trying to answer CNN host Dana Bash's questions about the Trump administration's long-forgotten promise of 90 trade deals during the president's 90-day pause on his sweeping reciprocal tariffs.

"We've seen three, and they're not concrete deals the way that we're used to seeing trade deals, they're frameworks," Bash said. Given Trump's reputation as a dealmaker, she asked: "Why haven't we seen the kind of deals he promised in the last 90 days?"

"Again, he didn't promise this," Bessent said. "And when we send out the hundred letters to these countries that will set their tariff rates ... so we're gonna have a hundred done in the next few days." Bessent was referring to Trump's Sunday night Truth Social post announcing he would notify multiple countries the following day of their new tariff rates.

"But that's not a deal, that's a threat," Bash pressed.
"No, that's the level. That's the deal. If you wanna trade with the United States, this is—" Bessent tried to explain.
"But that's not a negotiation, that's just a declaration," Bash said.
"Well, many of these countries never even contacted us," Bessent replied.

But just months ago, White House press secretary Karoline Leavitt claimed that the phones at the White House were "ringing off the hook" with countries calling to make deals with the United States. 

Another day, another lie.  The fun never stops, hey?  

Can't wait for Canada's new tariffs.

[HVC]

It'd be damning if not for the fact that most of the American public won't care.

[Josquius]

No, neither generative AI nor humans throw out what is most likely. Generative AI fabricates things and what it generates has no relationship to anything approaching a probability analysis.  People also, with rare exceptions of those of us who are trained to think in terms of probabilities (like DGuller), do not think about what is most probable. People generally try to make conclusions about what is the best or correct course of action. Something that generative AI programs are also incapable of doing.

AI absolutely does work that way. Based on its training data it guesses the most likely next outputs.

People....we may not be consciously aware of it but thats how we work too. Its all electrical neurons firing in our brains and outputs being created based on our training data. Its not a simple on or off as it is with computers. But fundamentally our brains are crunching possibilities and best courses of action based on them.

[crazy canuck]

Your post is a good example of a misconception people have about gender of AI and the misconception that it is doing anything which can be characterized as reasoning.

If generative AI analyzes the most probable outcome then why does it make so many mistakes and why does it outright fabricate information?

Generative AI is simply guessing at what the next word should be.  That is why it fabricates data and information.  If it was engaging in a probability analysis of the most likely outcome, as you assert, then it would not engaging in fabrication.

[Grey Fox]

It's not even words. LLMs are mainly segments based which closely align with our syllables.

[Josquius]

Your post is a good example of a misconception people have about gender of AI and the misconception that it is doing anything which can be characterized as reasoning.

If generative AI analyzes the most probable outcome then why does it make so many mistakes and why does it outright fabricate information?

Generative AI is simply guessing at what the next word should be.  That is why it fabricates data and information.  If it was engaging in a probability analysis of the most likely outcome, as you assert, then it would not engaging in fabrication.

An over simplification, it wouldn't do this with something so simple and rather for more complicated questions, but just for illustration....
If you ask AI what the population of Nowheresville is then it will try to find an answer for this.
If it can't find one it might nonetheless fill in a probable answer.

"The population of Nowheresville is 300" is a perfectly valid fill in the blank answer. It's possible it is true. Very possible it really isn't. It's what seems the most probable based on what the AI has access to - whether that's actual up to date census information or fuck all.

[crazy canuck]

It's not even words. LLMs are mainly segments based which closely align with our syllables.

Yes,

Your post is a good example of a misconception people have about gender of AI and the misconception that it is doing anything which can be characterized as reasoning.

If generative AI analyzes the most probable outcome then why does it make so many mistakes and why does it outright fabricate information?

Generative AI is simply guessing at what the next word should be.  That is why it fabricates data and information.  If it was engaging in a probability analysis of the most likely outcome, as you assert, then it would not engaging in fabrication.

An over simplification, it wouldn't do this with something so simple and rather for more complicated questions, but just for illustration....
If you ask AI what the population of Nowheresville is then it will try to find an answer for this.
If it can't find one it might nonetheless fill in a probable answer.

"The population of Nowheresville is 300" is a perfectly valid fill in the blank answer. It's possible it is true. Very possible it really isn't. It's what seems the most probable based on what the AI has access to - whether that's actual up to date census information or fuck all.

You are misusing the word valid

[grumbler]

It's not even words. LLMs are mainly segments based which closely align with our syllables.

Yes,

Your post is a good example of a misconception people have about gender of AI and the misconception that it is doing anything which can be characterized as reasoning.

If generative AI analyzes the most probable outcome then why does it make so many mistakes and why does it outright fabricate information?

Generative AI is simply guessing at what the next word should be.  That is why it fabricates data and information.  If it was engaging in a probability analysis of the most likely outcome, as you assert, then it would not engaging in fabrication.

An over simplification, it wouldn't do this with something so simple and rather for more complicated questions, but just for illustration....
If you ask AI what the population of Nowheresville is then it will try to find an answer for this.
If it can't find one it might nonetheless fill in a probable answer.

"The population of Nowheresville is 300" is a perfectly valid fill in the blank answer. It's possible it is true. Very possible it really isn't. It's what seems the most probable based on what the AI has access to - whether that's actual up to date census information or fuck all.

You are misusing the word valid

To be fair, you are misusing the word "guessing."

[Syt]

[Syt]

I feel this could be so much easier. Make being unemployed a crime. And since slavery is still allowed for convicted criminals ...

[Valmy]

On Medicaid? Do they think people can live on Medicaid and are just laying around getting those fat Medicaid checks?

Does anybody in this administration know what Medicaid is?