AI is outperforming humans in both IQ and creativity in 2021

[Dave]

he was teaching us HOW to study history in the process. That's the important thing.

It is, but if he was really teaching you that then you went to an exceptional school because you cannot study 'History' just from books and online sources, but need to get into original and contemporaneous documents. Many people think that everything is online now. It is extraordinary to me how much is online, but it is only touching the surface of the vast amount of data still held only in libraries and record offices, and that is why they are still vitally important to keep and protect.

The 'History' that I was taught at school was a still a boring regurgitation of facts and figures and I dropped it after GCSE. I hope that has changed and it does seem that it is taught as topics now (such as the 'history of medicine'.) I later got interested in my family history, and from that naturally came an interest in local history and social history, which in turn require some knowledge of the bigger picture to understand them. So, I went on to do an Advance Diploma.

I personally enjoy how technology has put a variety of factoids at my fingers.

So do I. When I was young, someone would ask a question, such as what is the highest mountain, and they'd never find out the answer. The Guinness Book of Records was started for that very reason. They might also ask what that song they heard was, and couldn't find that out. Now there is an App for that. I could wait for a bus forever without knowing if it was actually coming. There is an App for Train and Bus times, wherever you might be. There is an App for the local weather, the local tides, the stars visible in the sky. There is an App to identify what birdsong.you can hear It is totally mind-blowing what you can now find out at the touch of a button.

And the "AI" will have access to all of that too, but it is a parrot repeating things. It cannot make the decisons or interpretations that would enable it to make it's own content. It cannot visit a library or archive and find out something that has not been digitised yet.

Maybe a true AI will be able to do this soon. Maybe very soon. I think that will be much harder to achieve.

 

[OuttaInc]

Meh. Put me in the “unimpressed” category.

Those of us with children already know what it’s like to create something that thinks it’s smarter than us. Wake me when the ‘bots start wasting their days watching pointless TikTok videos and posting vids of themselves doing stupid stuff for likes.

 

[Wayne Mack]

I find the original article makes some pretty large leaps from the data presented to it's conclusion that "AI is outperforming humans in both IQ and creativity in 2021." I was not able to follow the logic from some (less than astounding) truths to that result.

"Artificial intelligence (AI) scores up to 40% higher than the average human on trivia. On SAT questions, AI scores 15% higher than an average college applicant." Anyone who has used internet search without AI in the past dozen years has been able to find the answers to trivia questions and other facts. To say that a well defined question about a well known truth can be found via the internet is not an indication of intelligence. This is merely repeating the things that these tools have been designed to do, parsing targeted queries and parsing the results. This does not seem to reflect intelligence.

“AI today is also prolifically creative, with one popular public model generating nearly as much new content as all 300 million Twitter users combined.” This seems to be equating quantity with creativity. Yes, computers can spew out text much faster than human typists. This says nothing about the text being generated.

'During another recorded session, Dr Thompson asks Leta to define the relationship between intelligence and communication. “Leta found a distinction immediately, saying: ‘Intelligence is about understanding the world, communication is about being understood by the world.’ A search of internet databases showed that sentence had never existed before being created by the AI in that moment.”' Again, this is what large language model systems were designed to do, to string together a sequence of words to make a likely sounding sentence. When the system goes beyond individual sentences, however, it has been widely observed that they loose all semblance of logic and reasoning.

Yes, technology has advanced, but using simplistic and somewhat non-mainstream definitions for intelligence and creativity is misleading. This is more a hype article than a discussion of current and projected AI capabilities.

 

[CrazyKB]

AI and IQ tests does not really tell us anything, except possibly to illustrate the limitations of IQ tests.

If I remember correctly, IQ tests (such as Stanford-Binet) were designed to illustrate who may require assistance in school. I don't think there's a definitive test of intelligence.

 

[Swank]

If I remember correctly, IQ tests (such as Stanford-Binet) were designed to illustrate who may require assistance in school. I don't think there's a definitive test of intelligence.

Show me someone that performs well on an IQ test who truly isn't intelligent.

 

[hitmouse]

Show me someone that performs well on an IQ test who truly isn't intelligent.

Define intelligence.

 

[Swank]

Define intelligence.

The ability to solve novel problems.

 

[HareBrain]

The ability to solve novel problems.

Well, that's me out -- mine are riddled with them.

 

[CrazyKB]

Define intelligence.

That's the problem, there really isn't a good definition of intelligence. And the old tests have been revamped in relation to changing views of what constitutes intelligence in Psychology. However, with that being said, modern intelligence tests do have merit.

 

[CrazyKB]

The ability to solve novel problems.

That's one component. I hold to the idea that human intelligence isn't 'one thing' but many. For example, during a discussion between Lawrence Krauss and Richard Dawkins, they spoke of their fascination with each of their respective fields. Lawrence wanted to study Biology, however soon realized that his peers (other students in his classes) were better than he was. He switched to Physics because he was better at it. While Dawkins stated he'd always loved Physics, however could never understand it on an academic level. Both of these people are highly intelligent, however can't process data equally well from different fields. Here's an article: A Harvard psychologist says humans have 8 types of intelligence. Which ones do you score the highest in?

 

[Swank]

That's one component. I hold to the idea that human intelligence isn't 'one thing' but many. For example, during a discussion between Lawrence Krauss and Richard Dawkins, they spoke of their fascination with each of their respective fields. Lawrence wanted to study Biology, however soon realized that his peers (other students in his classes) were better than he was. He switched to Physics because he was better at it. While Dawkins stated he'd always loved Physics, however could never understand it on an academic level. Both of these people are highly intelligent, however can't process data equally well from different fields. Here's an article: A Harvard psychologist says humans have 8 types of intelligence. Which ones do you score the highest in?

That's a bit like noting that Michael Jordan's professional baseball career wasn't as strong as his basketball career.

Being really smart guys, they realized what performing at the very peak of scientific pursuit would require. Most smart people wouldn't even see the problem as clearly as they did and have the self awareness to adjust accordingly. They would have probably been reasonable scientists in any field, just not truly great ones.

 

[Cthulhu.Science]

From Smithsonian Magazine - LINK

This A.I. Used Brain Scans to Recreate Images People Saw​

The technology, which was tested with four people, is still in its infancy but could one day help people communicate or decode dreams, researchers say.

The top row shows the actual images participants looked at, while the bottom row shows an A.I. recreation of each image based on the participant's brain scans. Edited from Takagi and Nishimoto / bioRxiv, 2022 under CC BY 4.0

Though it sounds like something out of a science fiction novel, researchers have successfully trained an artificial intelligence system to recreate images people have looked at based on their brain scans. The A.I. generated pictures of objects including a teddy bear, clock tower and airplane after participants viewed similar images.

--- Will this lead to the last image that dead people saw to help solve crime or find space criminals (Pitch Black).

 

[Swank]

It sounds like this method scans the electrical activity that goes along with seeing something, not necessarily remembering something past seen. Since the '70s we've known how to interact directly with the visual cortex. So I don't know if this is a leap forward (AI gets category and uses blurry data to presume an image), or if it is the predictable result of better scanning and very staid sample images.

I don't think what a person is looking at has much to do with what a person with zero brain activity once saw.

 

[Wayne Mack]

From Smithsonian Magazine - LINK

This A.I. Used Brain Scans to Recreate Images People Saw​

The technology, which was tested with four people, is still in its infancy but could one day help people communicate or decode dreams, researchers say.

The top row shows the actual images participants looked at, while the bottom row shows an A.I. recreation of each image based on the participant's brain scans. Edited from Takagi and Nishimoto / bioRxiv, 2022 under CC BY 4.0

Though it sounds like something out of a science fiction novel, researchers have successfully trained an artificial intelligence system to recreate images people have looked at based on their brain scans. The A.I. generated pictures of objects including a teddy bear, clock tower and airplane after participants viewed similar images.

--- Will this lead to the last image that dead people saw to help solve crime or find space criminals (Pitch Black).

Thanks for the link. This is interesting. I scanned through the article and what I understood of the process is that it:

• Train an AI to link MRI scans to viewed images
• Train an AI to link MRI scans of people viewing AI generated images to the generating text
• Takes an MRI scan of someone looking (not remembering) an image
• Use both the direct MRI scan to image and the MRI scan to text prompt to create a new AI generated image

Currently, the MRI scan to image only "recreate the perspective and layout that the participant had seen, but its generated images were of cloudy and nonspecific figures." The second part, "could recognize what object people were looking at by using the text descriptions from the training images." When these were combined, "if it received a brain scan that resembled one from its training marked as a person viewing an airplane, it would put an airplane into the generated image, following the perspective from the first model." The stated result was that "The technology achieved roughly 80 percent accuracy." It is unclear to me what 80% accuracy means in this context.

This is an interesting development, but I feel the headline is a bit of an exaggeration. The approach is not recreating the image seen as much as creating a facsimile of the image seen. It is more of an impression that a recreation.

 

[Christine Wheelwright]

I am very suspicious of this research. Or perhaps I am suspicious of the way it has been presented. I think you might just be able to tell whether someone is, say, looking at a square or a circle. Or a gun vs a flower. But the idea that you can recreate an image (like the bear) pixel by pixel, from measuring brain activity? Pure nonsense.

 

[CrazyKB]

That's a bit like noting that Michael Jordan's professional baseball career wasn't as strong as his basketball career.

Being really smart guys, they realized what performing at the very peak of scientific pursuit would require. Most smart people wouldn't even see the problem as clearly as they did and have the self awareness to adjust accordingly. They would have probably been reasonable scientists in any field, just not truly great ones.

In the case of Krauss, it was about 'preforming at the peak level of his field'. However, Dawkins has stated many times that he can't understand Physics the way other academics do. It's not that he wouldn't be an elite scientist in that field, but rather he couldn't be a scientist in that field period.

 

[Dave]

the idea that you can recreate an image (like the bear) pixel by pixel, from measuring brain activity? Pure nonsense.

I completely agree with your sentiments, but it isn't doing that really. It is recreating what the person remembered they had seen. So, that could be any teddy bear, tower with a clock, aeroplane or steam train from associated stock images. It is not a pixel-by-pixel recreation of the image.

 

[CrazyKB]

From Smithsonian Magazine - LINK

This A.I. Used Brain Scans to Recreate Images People Saw​

The technology, which was tested with four people, is still in its infancy but could one day help people communicate or decode dreams, researchers say.

The top row shows the actual images participants looked at, while the bottom row shows an A.I. recreation of each image based on the participant's brain scans. Edited from Takagi and Nishimoto / bioRxiv, 2022 under CC BY 4.0

Though it sounds like something out of a science fiction novel, researchers have successfully trained an artificial intelligence system to recreate images people have looked at based on their brain scans. The A.I. generated pictures of objects including a teddy bear, clock tower and airplane after participants viewed similar images.

--- Will this lead to the last image that dead people saw to help solve crime or find space criminals (Pitch Black).

In psychology there's what's known as the easy and hard problems. The easy problem is understanding how everything works (anything but easy), the hard problem is explaining how those processes produce our conscious experience, which may never be fully understood. Our visual system produces a simulation based on the data the brain receives (vision in this case), however what we perceive might only approximate what we're actually looking at. I don't understand how an AI can generate images based on brain activity, when we don't understand how the brain produces visual experience all that well, at least at this point. Not to mention, objects only have meaning if we have seen (and learned about) them before. For example, if you walk in a room your eyes will scan the room, this isn't a conscious activity, more like a reflex action (saccades). The brain is looking for objects it recognizes in an effort to understand the environment. And as far as being able to perceive what a dead person saw before they were murdered, well ... How could that be accomplished when the brain activity that was responsible wasn't recorded before death? I suppose you could employ some sort of Nano technology capable of doing so, then download the data, akin to 'Black Box' recorder in aircraft. However, if that's possible, I doubt it'll be a reality for a very long time.

 

[CrazyKB]

In psychology there's what's known as the easy and hard problems. The easy problem is understanding how everything works (anything but easy), the hard problem is explaining how those processes produce our conscious experiences, which may never be fully understood. Our visual system produces a simulation based on the data the brain receives (vision in this case), however what we perceive might only approximate what we're actually looking at. I don't understand how an AI can generate images based on brain activity, when we don't understand how the brain produces visual experiences all that well, at least at this point. Not to mention, objects only have meaning if we have seen (and learned about) them before. For example, if you walk in a room your eyes will scan the room, this isn't a conscious activity, more like a reflex action (saccades). The brain is looking for objects it recognizes in an effort to understand the environment. And as far as being able to perceive what a dead person saw before they were murdered, well ... How could that be accomplished when the brain activity that was responsible wasn't recorded before death? I suppose you could employ some sort of Nano technology capable of doing so, then download the data, akin to 'Black Box' recorder in aircraft. However, if that's possible, I doubt it'll be a reality for a very long time.

 

[Swank]

I completely agree with your sentiments, but it isn't doing that really. It is recreating what the person remembered they had seen. So, that could be any teddy bear, tower with a clock, aeroplane or steam train from associated stock images. It is not a pixel-by-pixel recreation of the image.

Do you know where it says "remembered"? I thought it was scanning what they were looking at while they were looking.