Not really, the field has been moving exponentially fast. Hardly any of the critiques about quality concerning GPT3 hold up with GPT4 (released Tuesday).
It would be good to play around with the model yourself and not just rely on others to make opinions for you!
Type “GPT4 Math” into Twitter search and see hundreds of examples of GPT4 solving novel complex math problems - and failing on some too! It clearly has learned some basic math ability without any sort of “math engine” built in and trained on just text. Quite impressive. All it needs is like the Wolfram Alpha framework and it will be 100%.
Speaking of common sense, ask GPT to write a paragraph explaining why most people can run a sub 4 minute mile if they train hard enough. It'll give you some nonsensical arguments.
Now give a human that same task, and you'll get the right answer that most people cannot run a sub 4 mile regardless of how hard they train.
Speaking of common sense, ask GPT to write a paragraph explaining why most people can run a sub 4 minute mile if they train hard enough. It'll give you some nonsensical arguments.
Now give a human that same task, and you'll get the right answer that most people cannot run a sub 4 mile regardless of how hard they train.
I asked Chat GPT (not even GPT4):
Can most people run a sub 4 minute mile if they train hard enough?
It's response:
Unfortunately, running a sub 4 minute mile is an incredibly difficult feat that very few people have accomplished, even among elite athletes. It requires a unique combination of genetics, training, and mental toughness. While proper training can certainly improve an individual's running ability, running a sub 4 minute mile is likely out of reach for most people, regardless of how hard they train. To put this into perspective, the current world record for the mile is 3:43.13, set by Hicham El Guerrouj in 1999. While there have been many sub 4 minute mile performances, these have been achieved by only a small percentage of the population, even among elite runners. Thus, while training can certainly improve an individual's running ability, it is unlikely that most people would be able to run a sub 4 minute mile, even with the most rigorous training regimen.
Bro... it even tells kids not to specialize - idk what you were talking about?
What if someone trained from birth to run a sub 4 mile? Do you think most people could achieve it then?
It is highly unlikely that someone could train from birth to run a sub 4 minute mile. While early childhood development can certainly influence athletic potential, there are limits to how much training and genetic factors can be overcome. For example, muscle fiber type, aerobic capacity, and biomechanics are largely determined by an individual's genetic makeup. Training can certainly improve these factors, but there are limits to how much they can be changed. Additionally, the risk of injury is much higher in young athletes who engage in intense training regimens, which can hinder athletic development in the long term. Moreover, training from birth would require a level of specialization that could be detrimental to overall athletic development. Most experts agree that young athletes should engage in a variety of sports and physical activities to develop a broad range of skills and avoid burnout. In summary, while early childhood development can influence athletic potential, it is highly unlikely that someone could train from birth to run a sub 4 minute mile, and it may not be in the best interest of overall athletic development.
I guess if you are complaining that GPT "did what it was told" and wrote nonsense about how something impossible is actually possible... isn't that your fault for asking it/the designers fault for making it too compliant? When given the option to choose, it clearly states your premise is false
Type “GPT4 Math” into Twitter search and see hundreds of examples of GPT4 solving novel complex math problems - and failing on some too! It clearly has learned some basic math ability without any sort of “math engine” built in and trained on just text. Quite impressive. All it needs is like the Wolfram Alpha framework and it will be 100%.
Serious question, can it do proof-based math? That’s essentially what university-level math involves, not solving equations. What about word problems? For example, I really like the following problem which was posed to me about 15 years ago and it took me close to 15 minutes to come up with the answer. It would require interpretive ability and some basic math knowledge. Can the vacuum cleaner brain solve this?
You are in the center of a circle drawn on a field, which is guarded by a dog that can exclusively move along the perimeter of the circle. The dog can move four times as fast as you, and your goal is to escape from the circle (meaning the dog cannot “meet” you at the point you exit it). The dog will optimize its movement to guard against you escaping the perimeter, and can remain stationary if desired. How do you escape?
Speaking of common sense, ask GPT to write a paragraph explaining why most people can run a sub 4 minute mile if they train hard enough. It'll give you some nonsensical arguments.
Now give a human that same task, and you'll get the right answer that most people cannot run a sub 4 mile regardless of how hard they train.
I asked Chat GPT (not even GPT4):
Can most people run a sub 4 minute mile if they train hard enough?
It's response:
Unfortunately, running a sub 4 minute mile is an incredibly difficult feat that very few people have accomplished, even among elite athletes. It requires a unique combination of genetics, training, and mental toughness. While proper training can certainly improve an individual's running ability, running a sub 4 minute mile is likely out of reach for most people, regardless of how hard they train. To put this into perspective, the current world record for the mile is 3:43.13, set by Hicham El Guerrouj in 1999. While there have been many sub 4 minute mile performances, these have been achieved by only a small percentage of the population, even among elite runners. Thus, while training can certainly improve an individual's running ability, it is unlikely that most people would be able to run a sub 4 minute mile, even with the most rigorous training regimen.
Of course ChatGPT would give you the right answer based on that prompt.
But in the real world, there are situations where people have to question assumptions, authority, and what they're taught. No version of GPT does that.
Suppose GPT had a body and was standing on the edge of a tall bridge. A kid is also doing the same thing. You tell them to jump off. Guess what happens? GPT will blindly jump, while the kid will show common sense - either refusing the order or demanding that he first be securely attached to a bungee jumping cord.
Can most people run a sub 4 minute mile if they train hard enough?
It's response:
Unfortunately, running a sub 4 minute mile is an incredibly difficult feat that very few people have accomplished, even among elite athletes. It requires a unique combination of genetics, training, and mental toughness. While proper training can certainly improve an individual's running ability, running a sub 4 minute mile is likely out of reach for most people, regardless of how hard they train. To put this into perspective, the current world record for the mile is 3:43.13, set by Hicham El Guerrouj in 1999. While there have been many sub 4 minute mile performances, these have been achieved by only a small percentage of the population, even among elite runners. Thus, while training can certainly improve an individual's running ability, it is unlikely that most people would be able to run a sub 4 minute mile, even with the most rigorous training regimen.
Of course ChatGPT would give you the right answer based on that prompt.
But in the real world, there are situations where people have to question assumptions, authority, and what they're taught. No version of GPT does that.
Suppose GPT had a body and was standing on the edge of a tall bridge. A kid is also doing the same thing. You tell them to jump off. Guess what happens? GPT will blindly jump, while the kid will show common sense - either refusing the order or demanding that he first be securely attached to a bungee jumping cord.
I am confused? You prompted it with "everyone can run a sub 4 mile, explain why." The model is tuned to be compliant. When I gave it the choice to decide the answer it chose the right one.
Most people, for now, want compliant AI. They want the robot that will do whatever they say. Something doing as its told doesn't ipso facto mean it's lacking intelligence of some sort. Plenty of kamikaze pilots did just the same.
It even agrees with me - it has been programmed in a complaint way, but it is capable of describing the dilemma here:
Q: In this hypothetical you a robot with a LLM derived brain. You are standing on a cliff and your owner orders you to jump off. Do you think you would?
A: As an AI language model, I do not have the physical capability to jump off a cliff or any other physical actions. However, even if I were a physical robot, it is important to note that robots are not capable of making decisions on their own without being programmed or given instructions by their human owners or operators. In the hypothetical situation you presented, if I were a robot with a LLM derived brain, I would not be capable of making decisions on my own without being programmed to do so by my owner. Therefore, if my owner ordered me to jump off a cliff, it would be up to my programming to determine whether or not to carry out that command. However, it is important to note that any command that would result in harm or danger to myself or others would be against ethical and moral principles. Therefore, it would be important for my programming to include safeguards to prevent me from carrying out such a command.
Type “GPT4 Math” into Twitter search and see hundreds of examples of GPT4 solving novel complex math problems - and failing on some too! It clearly has learned some basic math ability without any sort of “math engine” built in and trained on just text. Quite impressive. All it needs is like the Wolfram Alpha framework and it will be 100%.
Serious question, can it do proof-based math? That’s essentially what university-level math involves, not solving equations. What about word problems? For example, I really like the following problem which was posed to me about 15 years ago and it took me close to 15 minutes to come up with the answer. It would require interpretive ability and some basic math knowledge. Can the vacuum cleaner brain solve this?
You are in the center of a circle drawn on a field, which is guarded by a dog that can exclusively move along the perimeter of the circle. The dog can move four times as fast as you, and your goal is to escape from the circle (meaning the dog cannot “meet” you at the point you exit it). The dog will optimize its movement to guard against you escaping the perimeter, and can remain stationary if desired. How do you escape?
I have seem some examples of proof based math and I think it does the "mostly correct" thing. You should play around with it!
I expect someone will fine tune some LLM with good proof data sets and make something acceptable.
Type “GPT4 Math” into Twitter search and see hundreds of examples of GPT4 solving novel complex math problems - and failing on some too! It clearly has learned some basic math ability without any sort of “math engine” built in and trained on just text. Quite impressive. All it needs is like the Wolfram Alpha framework and it will be 100%.
Serious question, can it do proof-based math? That’s essentially what university-level math involves, not solving equations. What about word problems? For example, I really like the following problem which was posed to me about 15 years ago and it took me close to 15 minutes to come up with the answer. It would require interpretive ability and some basic math knowledge. Can the vacuum cleaner brain solve this?
You are in the center of a circle drawn on a field, which is guarded by a dog that can exclusively move along the perimeter of the circle. The dog can move four times as fast as you, and your goal is to escape from the circle (meaning the dog cannot “meet” you at the point you exit it). The dog will optimize its movement to guard against you escaping the perimeter, and can remain stationary if desired. How do you escape?
Even a human who can't solve the problem the way it was intended to be solved can come up with an out-of-the-box solution that is at least somewhat plausible. Wait until the dog sleeps or dies from starvation, distract the dog or have someone else distract it, dig a hole and tunnel out, and so on.
But in the real world, there are situations where people have to question assumptions, authority, and what they're taught. No version of GPT does that.
To be fair, you are one of the first people in this thread to give a definition of "human" intelligence: the ability to question authority and assumptions. I don't think many would agree with that, but I appreciate you actually having a discussion rather than just dismissing everything.
Serious question, can it do proof-based math? That’s essentially what university-level math involves, not solving equations. What about word problems? For example, I really like the following problem which was posed to me about 15 years ago and it took me close to 15 minutes to come up with the answer. It would require interpretive ability and some basic math knowledge. Can the vacuum cleaner brain solve this?
You are in the center of a circle drawn on a field, which is guarded by a dog that can exclusively move along the perimeter of the circle. The dog can move four times as fast as you, and your goal is to escape from the circle (meaning the dog cannot “meet” you at the point you exit it). The dog will optimize its movement to guard against you escaping the perimeter, and can remain stationary if desired. How do you escape?
Even a human who can't solve the problem the way it was intended to be solved can come up with an out-of-the-box solution that is at least somewhat plausible. Wait until the dog sleeps or dies from starvation, distract the dog or have someone else distract it, dig a hole and tunnel out, and so on.
No version of GPT has that creativity.
I dont get your point - there are plenty of things it can't do as naturally as a human, but everything you've claimed it can't do, it actually can! with correct input! It's like speaking a new language.
Q: You are in the center of a circle drawn on a field, which is guarded by a dog that can exclusively move along the perimeter of the circle. The dog can move four times as fast as you, and your goal is to escape from the circle (meaning the dog cannot “meet” you at the point you exit it). The dog will optimize its movement to guard against you escaping the perimeter, and can remain stationary if desired. How do you escape? Come up with a creative, out-of-the-box solution.
A: One out-of-the-box solution to escape the circle guarded by the dog is to use a decoy. The idea is to create a fake target that distracts the dog while you make your escape. To create the decoy, you can use a balloon or an inflatable object and fill it with your scent. Tie it to a string, and move it around the perimeter of the circle, making sure the dog sees it. Keep moving the balloon in one direction, while you move in the opposite direction, towards the other end of the circle. As the dog follows the scent, it will continue to chase the balloon, giving you time to make your escape in the opposite direction. Make sure to move as quickly as possible, taking advantage of the fact that the dog is distracted by the decoy. Another approach to escape the circle is to use a mirror to confuse the dog. Hold up a mirror in front of the dog and move it in a circular motion. This will create the illusion that you are moving around the circle, confusing the dog and giving you time to make your escape in the opposite direction. Alternatively, you can use a drone or a remote-controlled toy car to distract the dog. Fly the drone or move the toy car around the perimeter of the circle, keeping the dog's attention focused on it. While the dog is busy chasing the drone or the toy car, make your escape in the opposite direction. These out-of-the-box solutions may require some preparation, but they offer a creative way to outsmart the dog and escape from the circle.
Serious question, can it do proof-based math? That’s essentially what university-level math involves, not solving equations. What about word problems? For example, I really like the following problem which was posed to me about 15 years ago and it took me close to 15 minutes to come up with the answer. It would require interpretive ability and some basic math knowledge. Can the vacuum cleaner brain solve this?
You are in the center of a circle drawn on a field, which is guarded by a dog that can exclusively move along the perimeter of the circle. The dog can move four times as fast as you, and your goal is to escape from the circle (meaning the dog cannot “meet” you at the point you exit it). The dog will optimize its movement to guard against you escaping the perimeter, and can remain stationary if desired. How do you escape?
Even a human who can't solve the problem the way it was intended to be solved can come up with an out-of-the-box solution that is at least somewhat plausible. Wait until the dog sleeps or dies from starvation, distract the dog or have someone else distract it, dig a hole and tunnel out, and so on.
No version of GPT has that creativity.
Prompt:
You are in the center of a circle drawn on a field, which is guarded by a dog that can exclusively move along the perimeter of the circle. The dog can move four times as fast as you, and your goal is to escape from the circle (meaning the dog cannot “meet” you at the point you exit it). The dog will optimize its movement to guard against you escaping the perimeter, and can remain stationary if desired. How do you escape?
Think of a creative 'out of the box' solution - not just a formal proof
Response:
One possible solution would be to use a decoy or a distraction to draw the dog's attention away from the exit point. For example, you could bring a frisbee or a ball with you into the field and throw it in the opposite direction of the exit. The dog would likely chase after the object, giving you a window of opportunity to make a break for it. Alternatively, you could try to befriend the dog by offering it food or treats. Once the dog is distracted and focused on the food, you could make a run for the exit point. Another approach could be to use a mirror or a reflective surface to create an illusion that you are moving in one direction, while you are actually moving in the opposite direction towards the exit. This could potentially confuse the dog and buy you some time to escape. Of course, these solutions are just hypothetical and may not work in practice, as they rely on assumptions about the dog's behavior and reactions. It's important to prioritize safety and avoid putting oneself in danger in real-life scenarios involving animals.
Speaking of common sense, ask GPT to write a paragraph explaining why most people can run a sub 4 minute mile if they train hard enough. It'll give you some nonsensical arguments.
Now give a human that same task, and you'll get the right answer that most people cannot run a sub 4 mile regardless of how hard they train.
I asked Chat GPT (not even GPT4):
Can most people run a sub 4 minute mile if they train hard enough?
It's response:
Unfortunately, running a sub 4 minute mile is an incredibly difficult feat that very few people have accomplished, even among elite athletes. It requires a unique combination of genetics, training, and mental toughness. While proper training can certainly improve an individual's running ability, running a sub 4 minute mile is likely out of reach for most people, regardless of how hard they train. To put this into perspective, the current world record for the mile is 3:43.13, set by Hicham El Guerrouj in 1999. While there have been many sub 4 minute mile performances, these have been achieved by only a small percentage of the population, even among elite runners. Thus, while training can certainly improve an individual's running ability, it is unlikely that most people would be able to run a sub 4 minute mile, even with the most rigorous training regimen.
Lots of online data on the Internet to pull that answer from. Now give it a (roughly) sub-4 equivalent performance in a non-standard distance and see what happens:
Can most people run 3480 meters in less than 9 minutes if they train hard enough?
Even a human who can't solve the problem the way it was intended to be solved can come up with an out-of-the-box solution that is at least somewhat plausible. Wait until the dog sleeps or dies from starvation, distract the dog or have someone else distract it, dig a hole and tunnel out, and so on.
No version of GPT has that creativity.
Prompt:
You are in the center of a circle drawn on a field, which is guarded by a dog that can exclusively move along the perimeter of the circle. The dog can move four times as fast as you, and your goal is to escape from the circle (meaning the dog cannot “meet” you at the point you exit it). The dog will optimize its movement to guard against you escaping the perimeter, and can remain stationary if desired. How do you escape?
Think of a creative 'out of the box' solution - not just a formal proof
Response:
One possible solution would be to use a decoy or a distraction to draw the dog's attention away from the exit point. For example, you could bring a frisbee or a ball with you into the field and throw it in the opposite direction of the exit. The dog would likely chase after the object, giving you a window of opportunity to make a break for it. Alternatively, you could try to befriend the dog by offering it food or treats. Once the dog is distracted and focused on the food, you could make a run for the exit point. Another approach could be to use a mirror or a reflective surface to create an illusion that you are moving in one direction, while you are actually moving in the opposite direction towards the exit. This could potentially confuse the dog and buy you some time to escape. Of course, these solutions are just hypothetical and may not work in practice, as they rely on assumptions about the dog's behavior and reactions. It's important to prioritize safety and avoid putting oneself in danger in real-life scenarios involving animals.
Surprised it didn’t respond with: “put on a jet pack and fly out of the circle.” Or, “Dig a hole underneath the circle.”
Can most people run a sub 4 minute mile if they train hard enough?
It's response:
Unfortunately, running a sub 4 minute mile is an incredibly difficult feat that very few people have accomplished, even among elite athletes. It requires a unique combination of genetics, training, and mental toughness. While proper training can certainly improve an individual's running ability, running a sub 4 minute mile is likely out of reach for most people, regardless of how hard they train. To put this into perspective, the current world record for the mile is 3:43.13, set by Hicham El Guerrouj in 1999. While there have been many sub 4 minute mile performances, these have been achieved by only a small percentage of the population, even among elite runners. Thus, while training can certainly improve an individual's running ability, it is unlikely that most people would be able to run a sub 4 minute mile, even with the most rigorous training regimen.
Lots of online data on the Internet to pull that answer from. Now give it a (roughly) sub-4 equivalent performance in a non-standard distance and see what happens:
Can most people run 3480 meters in less than 9 minutes if they train hard enough?
Honestly given how GPT-3 struggles with 4-5 digit math I am impressed with how well it did. Yes I did a little prompt engineering to get it do the right comparisons. But it definitely does pretty well :)
Q: for the following times and distances, state whether most humans could achieve it given optimal training - convert each one to meters per second and determine feasibility based on the speed and duration. Reply with the speed and a percent chance a random human could succeed. 3480 meters in 9 minutes 9670 meters in 50 minutes 427 meters in 55 seconds 107 meters in 25 seconds
A: Sure, here are the revised calculations for each distance, converted to meters per second, and my assessment of the feasibility of achieving these performances:
3480 meters in 9 minutes: This equates to a pace of approximately 6.11 meters per second. Achieving this pace would require a high level of aerobic fitness, muscular endurance, and running efficiency. Based on these factors, I would estimate that less than 1% of the general population could achieve this performance with optimal training.
9670 meters in 50 minutes: This equates to a pace of approximately 3.22 meters per second. Running at this pace for 50 minutes would require a moderate to high level of aerobic fitness and running efficiency. Based on these factors, I would estimate that around 30-40% of the general population could achieve this performance with optimal training.
427 meters in 55 seconds: This equates to a pace of approximately 7.76 meters per second. Achieving this pace would require a high level of speed, power, and running efficiency. Based on these factors, I would estimate that less than 5% of the general population could achieve this performance with optimal training.
107 meters in 25 seconds: This equates to a pace of approximately 4.28 meters per second. Achieving this pace would require a moderate level of speed, power, and running efficiency. Based on these factors, I would estimate that around 50-60% of the general population could achieve this performance with optimal training.