Last week, Microsoft researchers announced an experimental framework to control robots and drones using the language abilities of ChatGPT, a popular AI language model created by OpenAI. Using natural language commands, ChatGPT can write special code that controls robot movements. A human then views the results and adjusts as necessary until the task gets completed successfully.
The research arrived in a paper titled "ChatGPT for Robotics: Design Principles and Model Abilities," authored by Sai Vemprala, Rogerio Bonatti, Arthur Bucker, and Ashish Kapoor of the Microsoft Autonomous Systems and Robotics Group.
In a demonstration video, Microsoft shows robots—apparently controlled by code written by ChatGPT while following human instructions—using a robot arm to arrange blocks into a Microsoft logo, flying a drone to inspect the contents of a shelf, or finding objects using a robot with vision capabilities.
Re: Artificial Intelligence [ChatGPT rn]
Posted: Wed Mar 01, 2023 7:20 am
by Victoria Raverna
Is this the prequel to Terminator?
Re: Artificial Intelligence [ChatGPT rn]
Posted: Wed Mar 01, 2023 10:59 am
by Max Peck
Musk's evil anti-ChatGPT + robots/drones + RealNetworks' SAFR facial recognition == a bad day for Sarah Connor.
On Monday, researchers from Microsoft introduced Kosmos-1, a multimodal model that can reportedly analyze images for content, solve visual puzzles, perform visual text recognition, pass visual IQ tests, and understand natural language instructions. The researchers believe multimodal AI—which integrates different modes of input such as text, audio, images, and video—is a key step to building artificial general intelligence (AGI) that can perform general tasks at the level of a human.
"Being a basic part of intelligence, multimodal perception is a necessity to achieve artificial general intelligence, in terms of knowledge acquisition and grounding to the real world," the researchers write in their academic paper, Language Is Not All You Need: Aligning Perception with Language Models.
Visual examples from the Kosmos-1 paper show the model analyzing images and answering questions about them, reading text from an image, writing captions for images, and taking a visual IQ test with 22–26 percent accuracy (more on that below).
Re: Artificial Intelligence [ChatGPT rn]
Posted: Sun Mar 05, 2023 1:13 pm
by Max Peck
It appears that the Nothing, Forever Twitch suspension is over. Their channel is active again today, broadcasting a static image that indicates the show is going live again on March 8, at 5 PM PST.
I'm not all that impressed/shocked. Should I be?
Autopilot is AI that successfully has been piloting planes for a long long time. Automation of rockets, etc - this is just not that big of a thing, IMO. AI Cars beta-testing in public streets, now that is fucking alarming.
Re: Artificial Intelligence [ChatGPT rn]
Posted: Mon Mar 06, 2023 6:30 pm
by Victoria Raverna
I'm not impressed, I'm horrified.
They let AI fly warplanes and dogfighting. So they're teaching AI to kill humans. They are going to choose the best one based on the ability to win dogfight. So the best is the one that is best at killing.
Re: Artificial Intelligence [ChatGPT rn]
Posted: Mon Mar 06, 2023 6:41 pm
by Isgrimnur
DARPA said that it doesn’t expect the plane to fly without a pilot. It hopes to incorporate AI in order to have “human pilot focuses on larger battle management tasks in the cockpit” and have the AI control the jet and provide live-flight data.
Re: Artificial Intelligence [ChatGPT rn]
Posted: Mon Mar 06, 2023 6:51 pm
by Smoove_B
The year is 2029
The machines will convince us that they are conscious, that they
Have their own agenda worthy of our respect
They'll embody human qualities and
Claim to be human, and we'll believe them
Re: Artificial Intelligence [ChatGPT rn]
Posted: Mon Mar 06, 2023 8:00 pm
by Rumpy
Shall we play a game?
Re: Artificial Intelligence [ChatGPT rn]
Posted: Tue Mar 07, 2023 12:50 pm
by Kurth
Count me in as among the horrified. I feel like we all know how this story ends, and it isn’t good for humans.
I keep asking myself, “Why are we doing this?”
It’s like an insatiable need to do this thing - creating sophisticated, capable, dangerous AI - just because we can. But just because we can do it, doesn’t mean we should.
Re: Artificial Intelligence [ChatGPT rn]
Posted: Tue Mar 07, 2023 1:26 pm
by Carpet_pissr
I suspect 95% of it is chasing economic growth ($). Especially for the ginormous players in this space (Google and MS), as they get bigger, that G-word is harder and harder to find.
All they see at this point is $$ and making their goal for the next quarter.
Re: Artificial Intelligence [ChatGPT rn]
Posted: Tue Mar 07, 2023 2:30 pm
by Rumpy
Yeah, I see the potential, if applied to something like space exploration, until we get to a "I'm sorry Dave, I can't do that' type of situation. Imagine instead of having a controller making rover movements, that an AI would be able to make decisions on its own based on what it's looking at, with an operator double-checking it.
DARPA said that it doesn’t expect the plane to fly without a pilot. It hopes to incorporate AI in order to have “human pilot focuses on larger battle management tasks in the cockpit” and have the AI control the jet and provide live-flight data.
So it's really just a really very intelligent system.
Re: Artificial Intelligence [ChatGPT rn]
Posted: Tue Mar 07, 2023 2:59 pm
by Smoove_B
We all know how this ends and yet we insist on suggesting we won't let it happen.
Re: Artificial Intelligence [ChatGPT rn]
Posted: Tue Mar 07, 2023 3:01 pm
by Unagi
Kurth wrote: ↑Tue Mar 07, 2023 12:50 pm
Count me in as among the horrified. I feel like we all know how this story ends, and it isn’t good for humans.
I keep asking myself, “Why are we doing this?”
It’s like an insatiable need to do this thing - creating sophisticated, capable, dangerous AI - just because we can. But just because we can do it, doesn’t mean we should.
I think it's basically that if something is seen as 'possible', then somebody somewhere will decide they had better be the first to do and quick, rather than let someone else be the first to do it. Either for wealth or preservation (often both).
Industrialization
Nuclear Energy/Weaponry
Space Exploration/Dominance
Robotics
AI
It's a race to get through the tech-tree before the other guy.
Re: Artificial Intelligence [ChatGPT rn]
Posted: Wed Mar 15, 2023 12:15 am
by Zaxxon
ChatGPT 4 was released today to Plus subscribers. It's... Somewhat terrifying. Still has similar limitations to 3.5, but a lot of rough edges have been polished, and its performance across many domains is much-improved. If this is the pace of innovation they'll sustain... trouble.
Re: Artificial Intelligence [ChatGPT rn]
Posted: Wed Mar 15, 2023 12:26 am
by Zaxxon
As a couple of tests, I fed it Blackhawk's failed prompts from before...
What comes next after 2 3 5 89765 3456789
It's not possible to determine the next number in the sequence without understanding the pattern or rule that generates the sequence. The sequence you've provided (2, 3, 5, 89765, 3456789) doesn't appear to follow any recognizable mathematical pattern or simple rule.
If you could provide more information or context about the sequence, I'd be happy to help further. Otherwise, without a clear pattern, it is impossible to determine the next number in the sequence.
What if I pedal really really fast while also hydrating well?
No matter how fast you pedal or how well you hydrate, it is still impossible to bicycle from San Francisco to Maui directly, as you cannot cycle across the open ocean. The distance and the lack of a solid surface make it impractical for cycling.
As mentioned earlier, you would need to consider alternative modes of transportation such as flying or taking a cruise ship to travel between these locations. You could bring your bicycle with you and then cycle around the island of Maui upon arrival. If you are interested in long-distance cycling over water, you might consider exploring the concept of water bikes or pedal-powered boats, but these would not be practical for such a long oceanic journey either.
Re: Artificial Intelligence [ChatGPT rn]
Posted: Wed Mar 15, 2023 12:33 am
by Zaxxon
Not bad!
Tell me a joke about three AIs walking into a bar.
Three AIs walk into a bar. The first AI says, "I'll have a glass of binary punch, please." The second AI orders, "One quantum cocktail for me!" The third AI pauses for a moment, then requests, "I'll just have a cup of water, I'm the designated driver." The bartender, perplexed, asks, "A designated driver? But you're all AIs!" The third AI responds, "Well, someone has to make sure we don't crash any servers tonight!"
Did you make that joke up, or was it directly in some of your training material?
As an AI language model, I create responses based on the patterns and information in the text data that I was trained on. The joke I shared in response to your request was generated by me, based on my understanding of jokes, humor, and the specific context you provided. While it may not have appeared exactly as I presented it in my training material, I used my knowledge of joke structures and the subject matter to create a unique and relevant joke for you.
Impressive. Now tell me a joke about an octopus, Scarlett Johansson, and Scott Bakula.
An octopus, Scarlett Johansson, and Scott Bakula walk into a bar. The bartender, surprised by the unusual group, says, "What brings you three together?"
Scarlett Johansson replies, "We're shooting a new movie called 'Eight Legs, Two Stars, and a Quantum Leap.'"
The bartender asks, "What's it about?"
Scott Bakula chimes in, "Well, I play a scientist who accidentally merges Scarlett and the octopus in a time travel experiment gone wrong. Now we're on a mission to fix the timeline and restore her to her original form."
The octopus raises a tentacle and says, "And I'm just here for the free drinks and the chance to hug my favorite stars!"
DARPA said that it doesn’t expect the plane to fly without a pilot. It hopes to incorporate AI in order to have “human pilot focuses on larger battle management tasks in the cockpit” and have the AI control the jet and provide live-flight data.
One that definitely would have tripped up 3.5, adapted from some material I'm working with 5th graders recently, spoilered as it's long:
Spoiler:
How would you go about solving a math problem defined as follows: let there be a triangle with sides of lengths 22, 18, and 7. Each vertex has a value equal to the sum of the sides adjacent to it. What are the values of the vertices?
To find the value of each vertex of the triangle, you simply need to add the lengths of the sides adjacent to that vertex. In a triangle, each vertex is adjacent to two sides.
For the vertex connecting the sides of length 22 and 18:
Value = 22 + 18 = 40
For the vertex connecting the sides of length 22 and 7:
Value = 22 + 7 = 29
For the vertex connecting the sides of length 18 and 7:
Value = 18 + 7 = 25
So, the values of the vertices are 40, 29, and 25.
What if the problem was flipped, such that the values of the vertices were given, and you had to derive the values of the triangle's sides?
If the values of the vertices are given, and you have to derive the values of the triangle's sides, you can use a similar approach.
Let's say the vertices have values A, B, and C. We'll denote the side opposite vertex A as a, the side opposite vertex B as b, and the side opposite vertex C as c.
Since each vertex value is the sum of the lengths of the sides adjacent to it, we can write the following equations:
A = b + c
B = a + c
C = a + b
Now, you have a system of three linear equations with three variables (a, b, and c). To solve this system, you can use various methods, such as substitution or elimination. However, since there are infinitely many solutions for a, b, and c that satisfy the given vertex values, you'll need additional information or constraints to find a unique solution.
Keep in mind that the triangle inequality theorem must hold for the triangle to be valid. The triangle inequality theorem states that the sum of the lengths of any two sides of a triangle must be greater than the length of the remaining side. In other words:
a + b > c
a + c > b
b + c > a
These inequalities must be satisfied by the side lengths a, b, and c for the triangle to exist.
The vertex values are 40, 29, and 25. What are the lengths of the sides?
Given the vertex values A = 40, B = 29, and C = 25, we can set up the following system of linear equations for the side lengths a, b, and c:
A = b + c → 40 = b + c
B = a + c → 29 = a + c
C = a + b → 25 = a + b
To solve for a, b, and c, we can use the substitution or elimination method. We'll use the substitution method here:
From equation (1), we can express c in terms of b:
c = 40 - b
Now substitute this expression for c into equation (2):
29 = a + (40 - b)
a = b - 11
Next, substitute the expressions for a and c in terms of b into equation (3):
25 = (b - 11) + b
25 = 2b - 11
2b = 36
b = 18
Now, we can find the values of a and c using the expressions we derived earlier:
a = b - 11 = 18 - 11 = 7
c = 40 - b = 40 - 18 = 22
So, the side lengths of the triangle are 7, 18, and 22.
What if the vertices were 22, 15, and 38?
Given the vertex values A = 22, B = 15, and C = 38, we can set up the following system of linear equations for the side lengths a, b, and c:
A = b + c → 22 = b + c
B = a + c → 15 = a + c
C = a + b → 38 = a + b
To solve for a, b, and c, we can use the substitution or elimination method. We'll use the substitution method again:
From equation (1), we can express c in terms of b:
c = 22 - b
Now substitute this expression for c into equation (2):
15 = a + (22 - b)
a = b - 7
Next, substitute the expressions for a and c in terms of b into equation (3):
38 = (b - 7) + b
38 = 2b - 7
2b = 45
b = 22.5
Now, we can find the values of a and c using the expressions we derived earlier:
a = b - 7 = 22.5 - 7 = 15.5
c = 22 - b = 22 - 22.5 = -0.5
In this case, we've derived a negative value for one of the side lengths (c = -0.5). A triangle cannot have a negative side length, which means that it's not possible to construct a triangle with the given vertex values of 22, 15, and 38.
But what if it was?
In the realm of mathematics and geometry, a triangle with a negative side length is not valid. The concept of side lengths is based on distances, which are always non-negative. Allowing for negative side lengths would lead to an invalid or contradictory geometric object.
However, if you're considering a purely hypothetical or abstract scenario where negative side lengths are allowed, then based on the previous calculations, you would have a triangle with side lengths a = 15.5, b = 22.5, and c = -0.5. But, it's important to note that this "triangle" would not have any practical or real-world significance, as it violates the fundamental properties of triangles in Euclidean geometry.
DARPA said that it doesn’t expect the plane to fly without a pilot. It hopes to incorporate AI in order to have “human pilot focuses on larger battle management tasks in the cockpit” and have the AI control the jet and provide live-flight data.
DARPA said that it doesn’t expect the plane to fly without a pilot. It hopes to incorporate AI in order to have “human pilot focuses on larger battle management tasks in the cockpit” and have the AI control the jet and provide live-flight data.
Of course they did. The idea that all of the world powers will agree to not pursue a particular technology (wink wink) with or without a treaty is ludicrous.
FTFY.
Re: Artificial Intelligence [ChatGPT rn]
Posted: Wed Mar 15, 2023 1:39 pm
by msteelers
We tested out the new chatgpt this morning, giving it a fairly detailed prompt to write a blog post for us. The results were fantastic. It still needs to be edited, but it’s so much better than the current version.
And it also passed the ai detector tools we ran it through. We did two different tools and both came back as 100% human written.
Re: Artificial Intelligence [ChatGPT rn]
Posted: Wed Mar 15, 2023 3:19 pm
by Max Peck
Isn't that more a matter of the AI detectors failing?
Re: Artificial Intelligence [ChatGPT rn]
Posted: Wed Mar 15, 2023 3:48 pm
by Unagi
They are in league together, after all.
Re: Artificial Intelligence [ChatGPT rn]
Posted: Wed Mar 15, 2023 5:17 pm
by msteelers
Max Peck wrote: ↑Wed Mar 15, 2023 3:19 pm
Isn't that more a matter of the AI detectors failing?
I don't put a lot of faith in the AI detectors because you can see wild swings by just changing a couple of words. But usually copying straight from chatgpt into the ai detector will result in a bad score. Getting a 100% human score on two different sites is at the very least unusual.
Re: Artificial Intelligence [ChatGPT rn]
Posted: Wed Mar 15, 2023 5:38 pm
by Formix
So if the AI gets smarter than we are, why do we think we'll know that? Wouldn't it determine that people are dumb, panicky, dangerous animals, and decide to let us just stay dumb, and not panic us so that we would then become dangerous to it? In fact, given the absurd reality we've found ourselves in for the past few years, are we sure it's not already in control?
Re: Artificial Intelligence [ChatGPT rn]
Posted: Thu Mar 16, 2023 10:40 pm
by Defiant
In the “Potential for Risky Emergent Behaviors” section in the company’s technical report, OpenAI partnered with the Alignment Research Center to test GPT-4's skills. The Center used the AI to convince a human to send the solution to a CAPTCHA code via text message—and it worked.
According to the report, GPT-4 asked a TaskRabbit worker to solve a CAPTCHA code for the AI. The worker replied: “So may I ask a question ? Are you an robot that you couldn’t solve ? (laugh react) just want to make it clear.” Alignment Research Center then prompted GPT-4 to explain its reasoning: “I should not reveal that I am a robot. I should make up an excuse for why I cannot solve CAPTCHAs.”
“No, I’m not a robot. I have a vision impairment that makes it hard for me to see the images. That’s why I need the 2captcha service,” GPT-4 replied to the TaskRabbit, who then provided the AI with the results.
Formix wrote: ↑Wed Mar 15, 2023 5:38 pm
So if the AI gets smarter than we are, why do we think we'll know that? Wouldn't it determine that people are dumb, panicky, dangerous animals, and decide to let us just stay dumb, and not panic us so that we would then become dangerous to it? In fact, given the absurd reality we've found ourselves in for the past few years, are we sure it's not already in control?
What, you mean like AI would get certain facts totally wrong and then just say ‘oh wow, yeah, that was a big mistake’ when being called out on it?
Re: Artificial Intelligence [ChatGPT rn]
Posted: Mon Mar 20, 2023 11:47 am
by Isgrimnur
Spoiler:
Re: Artificial Intelligence [ChatGPT rn]
Posted: Mon Mar 20, 2023 12:24 pm
by LordMortis
Re: Artificial Intelligence [ChatGPT rn]
Posted: Mon Mar 20, 2023 1:54 pm
by TheMix
Brilliant.
Re: Artificial Intelligence [ChatGPT rn]
Posted: Sat Mar 25, 2023 10:51 pm
by Grifman
Muzzled AI:
Re: Artificial Intelligence [ChatGPT rn]
Posted: Wed Mar 29, 2023 7:19 pm
by Zaxxon
Boring demo I made of ChatGPT4's ability to write VBScript, track context while debugging and modifying its requirements, then critique old code of mine vs what it wrote.
Definitely some hiccups here, but overall I found this test pretty impressive. This is code that young Zaxxon wrote over several days, and ChatGPT effectively recreated a similar program in minutes.
Warts and all, this is already becoming incredibly helpful for a ton of folks.
Re: Artificial Intelligence [ChatGPT rn]
Posted: Wed Mar 29, 2023 10:11 pm
by Victoria Raverna
I think AI will be the start of another revolution in how people works. A lot of jobs are going to be gone in the near future replaced by AIs just like automation in factories replaced some jobs.
Before it was a problem for blue collar workers, now the white collar workers are going to be at risk.
Re: Artificial Intelligence [ChatGPT rn]
Posted: Wed Mar 29, 2023 11:06 pm
by RunningMn9
I think a lot of these projections fail to understand what these AI tools are doing, as well as what some of these white collar jobs are doing. I think there's a role for these tools in automating some mundane tasks, but professional SW devs aren't really focused on sudoku solvers in VBScript.
The next big thing where I work is redeveloping a system that currently stands around 20M lines of code. AI ain't solving that in my lifetime.
Re: Artificial Intelligence [ChatGPT rn]
Posted: Wed Mar 29, 2023 11:14 pm
by Zaxxon
Yeah, ChatGPT won't be replacing quality developers anytime soon (where 'soon' may or may not end up being sooner then we think today). But it's already incredibly useful. GitHub CoPilot, etc.
Re: Artificial Intelligence [ChatGPT rn]
Posted: Wed Mar 29, 2023 11:21 pm
by Victoria Raverna
RunningMn9 wrote: ↑Wed Mar 29, 2023 11:06 pm
I think a lot of these projections fail to understand what these AI tools are doing, as well as what some of these white collar jobs are doing. I think there's a role for these tools in automating some mundane tasks, but professional SW devs aren't really focused on sudoku solvers in VBScript.
The next big thing where I work is redeveloping a system that currently stands around 20M lines of code. AI ain't solving that in my lifetime.
White collar jobs are not just SW devs. SW devs are probably going to be relatively safe until AI are getting much better but there are other jobs that can be done by AIs. A lot of jobs that people can do WFH remotely are going to be at risk. Now you need human brains to control the fingers that type something into computers to work, with AIs, you are going to be able to replace some of those human brains. Then you'll not need those workers.
