- 2,156
- United Kingdom
Full AI - The End of Humanity?
- Thread starter TenEightyOne
- 415 comments
- 32,503 views
- 19,956
- NJ/USA
- Blitzbay
- Blitzbay
This must be the new math they are talking about.
- 15,313
- TRAPPIST-1g
- ProjectWHaT
OpenAI just announced ChatGPT plugins
This essentially allows ChatGPT to interact with third party services purely using natural language to guide ChatGPT to do what you want.
There's already a plugin for Wolfram Alpha
Using Wolfram Alpha, ChatGPT is able to correctly answer all "nontrivial calculations" that is would fail at before
Before, you had to have specific knowledge and expertise to know what to look up and the tools needed to get what you want. Now, using natural language, you can get the same result with with zero prior knowledge or skill in fraction of the time
Here's a comparison between using just ChaptGPT and with using Wolfram Alpha
At the end of the blog, Wolfram briefly discusses his history with developing Wolfram and his outlook of the future
From Tim Urban's article on AI back from 2015 (which I highly recommend reading, especially now: https://waitbutwhy.com/2015/01/artificial-intelligence-revolution-1.html [language warning]), he drew these diagrams
It seems like we've reached a point with ChatGPT where progress has jumped exponentially in the past few years. GPT-3 was released in 2020. DALL-E 2 and GPT-3.5 was released in 2022. Bing Chat was released released last month, giving ChatGPT access to the internet. GPT-4 was just released a week ago. They've now announced plug-ins today allowing it to work with third party interfaces.
As Wolfram noted in his article, all of today's technologies were pioneered back in the 40s and 50s when these statistical models like the perceptron and neural networks were first conceptualized. It has taken these 80 years for the processing power and data to catch up to make all of this possible.
There's a lot of other good commentary on these Hacker News posts
news.ycombinator.com
news.ycombinator.com
Here's a comment I like:
There's a research paper by a bunch of Microsoft engineers that evaluate GPT-4's capabilities as an AGI which is titled "Sparks of Artificial General Intelligence: Early experiments with GPT-4". It's 95 pages long, excluding references and appendices. I plan to read some of it this week, but here is what the abstract says:
ChatGPT plugins
We’ve implemented initial support for plugins in ChatGPT. Plugins are tools designed specifically for language models with safety as a core principle, and help ChatGPT access up-to-date information, run computations, or use third-party services.
openai.com
This essentially allows ChatGPT to interact with third party services purely using natural language to guide ChatGPT to do what you want.
There's already a plugin for Wolfram Alpha
ChatGPT Gets Its “Wolfram Superpowers”!
Wolfram plugin gives computationally accurate answers to ChatGPT queries using Wolfram's data knowledgebase and language. Custom visualizations are given as well. Stephen Wolfram explains how it works.
writings.stephenwolfram.com
Using Wolfram Alpha, ChatGPT is able to correctly answer all "nontrivial calculations" that is would fail at before
Before, you had to have specific knowledge and expertise to know what to look up and the tools needed to get what you want. Now, using natural language, you can get the same result with with zero prior knowledge or skill in fraction of the time
Here's a comparison between using just ChaptGPT and with using Wolfram Alpha
At the end of the blog, Wolfram briefly discusses his history with developing Wolfram and his outlook of the future
From Tim Urban's article on AI back from 2015 (which I highly recommend reading, especially now: https://waitbutwhy.com/2015/01/artificial-intelligence-revolution-1.html [language warning]), he drew these diagrams
It seems like we've reached a point with ChatGPT where progress has jumped exponentially in the past few years. GPT-3 was released in 2020. DALL-E 2 and GPT-3.5 was released in 2022. Bing Chat was released released last month, giving ChatGPT access to the internet. GPT-4 was just released a week ago. They've now announced plug-ins today allowing it to work with third party interfaces.
As Wolfram noted in his article, all of today's technologies were pioneered back in the 40s and 50s when these statistical models like the perceptron and neural networks were first conceptualized. It has taken these 80 years for the processing power and data to catch up to make all of this possible.
There's a lot of other good commentary on these Hacker News posts
ChatGPT Plugins | Hacker News
ChatGPT can now call Wolfram Alpha | Hacker News
Here's a comment I like:
There's a research paper by a bunch of Microsoft engineers that evaluate GPT-4's capabilities as an AGI which is titled "Sparks of Artificial General Intelligence: Early experiments with GPT-4". It's 95 pages long, excluding references and appendices. I plan to read some of it this week, but here is what the abstract says:
- 7,573
- Horbury, West Yorkshire
- GTP_Sprite
I’m intrigued by all this stuff, but I can’t wrap my small brain around a lot of it.
At what point will any of this become useful to someone like myself (layman)? and in what form? There has to be a point where it’s filtered into something an everyday person uses but doesn’t realise and takes it for granted.
Take touchscreen phones. Back when I registered for GT Planet I had an old Nokia. It had a monochrome screen, basic keyboard, no camera and the height of its fancy features was you could program your own ringtone and change the outer front cover for something new.
Touchscreens, multi megapixel cameras, gigabytes of storage and gaming on a small device you could put in your pocket seemed like an age away. Now we all take it for granted and any phone which doesn’t have these features or improve upon them is seen as inferior.
At what point do general AI of this scale become common place and in what form? Serious question, I have no idea and really can’t see what benefit to me it all is, yet.
At what point will any of this become useful to someone like myself (layman)? and in what form? There has to be a point where it’s filtered into something an everyday person uses but doesn’t realise and takes it for granted.
Take touchscreen phones. Back when I registered for GT Planet I had an old Nokia. It had a monochrome screen, basic keyboard, no camera and the height of its fancy features was you could program your own ringtone and change the outer front cover for something new.
Touchscreens, multi megapixel cameras, gigabytes of storage and gaming on a small device you could put in your pocket seemed like an age away. Now we all take it for granted and any phone which doesn’t have these features or improve upon them is seen as inferior.
At what point do general AI of this scale become common place and in what form? Serious question, I have no idea and really can’t see what benefit to me it all is, yet.
Last edited:
- 15,313
- TRAPPIST-1g
- ProjectWHaT
I think it can be useful right now. The beauty, and horror, of ChatGPT is that it works with natural language. You can write to it like how you can write to it like another human, with the same slang and words you use everyday, without having to rephrase things and jump through hoops in specific ways for a computer to understand.
A very basic example people use, and is mentioned in Tom Scott's video (which is very good and highly recommended), is writing emails. You can prompt ChatGPT about what sort of email you would like to write. It can be formal, informal, etc. and within seconds it will spit out paragraphs that you can quickly proofread and send.
Or, you can prompt ChatGPT with a complex article and have it summarize it for you. You don't have to learn any background knowledge. You can chat with ChatGPT as it walks you through the complex article.
(Side note: these AI systems, at the moment, can be confidently incorrect about a lot of things. You definitely should do your own research when asking it high risk questions such as medical, legal, financial advice, etc.)
At the same time, ChatGPT can and has been integrated into things people use everyday. Some of the headlining companies are Microsoft, Goldman Sachs, Slack, Coca Cola, Duolingo, Shopify, Discord, etc. A lot of these are still quite rudimentary integrations.
I think we're so early on on the exponential progress curve that we truly don't know exactly how common place this will become and in what forms of how we will be using it. At the moment, we're using ChatGPT for quite simple things, asking it trivial questions, summarizing articles, writing emails, debugging code, etc. We don't actually know what its true capabilities are. We don't know how far we can take it. We don't know anymore what the future could look like with AI.
As Tom Scott said in his video, this is similar to the beginnings of the internet and Napster. We see the new technology and we know it can be highly capable, but at this point, we truly don't know what we can do with it. We can only see in retrospect where we were and where we were heading
Out of curiosity, without doxing yourself or revealing too much, could you say how are you using AI at work?
- 2,156
- United Kingdom
Help write South Park episodes (maybe)
This is definitely true.
While it can pass the USMLE, it does get medical information wrong that any duly dilligent doctor should know.
Last edited:
Danoff
Premium
- 32,991
- Mile High City
Basically it's a search algorithm that uses AI to generate good search results without a specific query. Think kinda like google image searching using a seed image.
Danoff
Premium
- 32,991
- Mile High City
A lot of people spell out doom and gloom over AI development. In some cases, because AI will take our jerbs, and in other cases, because AI will become sentient and attack. I suppose there's yet another category of concern that is centered around poorly programmed AI doing something awful and being very difficult to stop. I'm actually not super worried about any of those things, and I have laid out reasons for each in this thread.
It occurred to me that there is another threat, and one which I've only just started seriously considering. A rapid increase in technological development.
You might think I'm going to say that this is a problem because culturally we can't adapt culturally, or because it'll be unequal, or something like that. Nope. I'm mostly concerned with what we'll do with the rapid increase in technology. With AI, we could potentially engineer a super virus, or find a way to fundamentally alter spacetime in a cascade (it is possible) and end the entire universe.
Perhaps another thing to be concerned about is the Beggars in Spain scenario (good book, worth a read). Suppose we turn AI on genetic engineering of our species. We may be able to very rapidly do things like eliminate the need for sleep, or eliminate aging. Perhaps we could fundamentally improve the functioning of the brain, memory, or the immune system. Now that might sound amazing. In fact, that might sound exactly like the kind of thing we should turn AI onto solving. But consider what happens when it's not you who benefits. Consider what happens when it's a new generation of superhumans born from synthetic DNA that can't be offered to already existing humans. You and I could find ourselves considered to be an "inferior" species altogether. We've seen, in the past, what happens to even other humans who are considered inherently inferior in some way. I would definitely feel unsettled being surrounded by a new generation of immortal, super smart, untiring superhumans. It'd be great for them, I just hope they're nice.
I'm sure there are many other ways that AI could bring us some paradigm altering technology that we end up ruining our lives with.
It occurred to me that there is another threat, and one which I've only just started seriously considering. A rapid increase in technological development.
You might think I'm going to say that this is a problem because culturally we can't adapt culturally, or because it'll be unequal, or something like that. Nope. I'm mostly concerned with what we'll do with the rapid increase in technology. With AI, we could potentially engineer a super virus, or find a way to fundamentally alter spacetime in a cascade (it is possible) and end the entire universe.
Perhaps another thing to be concerned about is the Beggars in Spain scenario (good book, worth a read). Suppose we turn AI on genetic engineering of our species. We may be able to very rapidly do things like eliminate the need for sleep, or eliminate aging. Perhaps we could fundamentally improve the functioning of the brain, memory, or the immune system. Now that might sound amazing. In fact, that might sound exactly like the kind of thing we should turn AI onto solving. But consider what happens when it's not you who benefits. Consider what happens when it's a new generation of superhumans born from synthetic DNA that can't be offered to already existing humans. You and I could find ourselves considered to be an "inferior" species altogether. We've seen, in the past, what happens to even other humans who are considered inherently inferior in some way. I would definitely feel unsettled being surrounded by a new generation of immortal, super smart, untiring superhumans. It'd be great for them, I just hope they're nice.
I'm sure there are many other ways that AI could bring us some paradigm altering technology that we end up ruining our lives with.
Last edited:
- 7,573
- Horbury, West Yorkshire
- GTP_Sprite
Without turning to SciFi too much, but the Eugenics Wars springs to mind. I agree AI would be a good candidate for speeding up genetic modification / alteration.
I’m more intrigued to see what AI can do in other areas like designing green solutions to energy, transport and space travel.
If eventually AI does help us solve some of these things (especially energy) then I’m all for it.
My worry is AI in the wrong hands and used to wage war. While humans using cyber attacks can wreak havoc already, an AI would be devastating.
- 4,819
- Kentucky
- barrett5
- BarrettMopar
Someone connected their brain to GPT 4.
I guess now that GPT 4 allows for plugins that access the internet it could be considered as the first way people can become a robot. I haven't tried the bing version of GPT 4 yet, but apparently it can create websites so I guess people at the very least can create websites with their mind. It would be cool if all this can be used with Midjourney to visualize dreams and maybe someday people can livestream their dreams.
I guess now that GPT 4 allows for plugins that access the internet it could be considered as the first way people can become a robot. I haven't tried the bing version of GPT 4 yet, but apparently it can create websites so I guess people at the very least can create websites with their mind. It would be cool if all this can be used with Midjourney to visualize dreams and maybe someday people can livestream their dreams.
- 2,156
- United Kingdom
- 11,319
- Marin County
- 15,313
- TRAPPIST-1g
- ProjectWHaT
Here are some generative AI use cases in game development
ROAD_DOGG33J
Premium
- 13,896
- IL, USA
- holyc0w1
- holyc0w
The newest version of ChatGPT passed the US medical licensing exam with flying colors — and diagnosed a 1 in 100,000 condition in seconds
ChatGPT's latest software upgrade, called GPT-4, is "better than many doctors I've observed" at clinical diagnosis, one physician said.
www.insider.com
- 2,156
- United Kingdom
AI ‘humans’ in virtual world plot and gossip like us
Isabella is throwing a party and decides to ask her friend Maria to help put up decorations. Maria agrees and sees an opportunity — she invites Klaus, her secre
www.thetimes.co.uk
- 1,274
- ^mfw that Midget II
- Silhouette Drops
I don't really delve in messing around with AI programs like chatGPT, but I have been thinking about how advanced AI is getting.
granted, this is a dream, but how real it was caused me to snap back into reality.
Imagine this: AI becomes so advanced, that we'll have something akin to VEGA from DOOM 2016 and Eternal. Silly, I know, but think about it. Within the next decade, maybe even sooner, we'll have AI programs capable of using voice that are so far advanced, the line between real and fake will literally become nonexistent bar you're involved from development of said program.
To summarize my dream in a few short sentences, it's like a twilight zone episode. Friend called me up whom I haven't talked to since high school. Did some catching up. Wanted to hang out, but kept getting put on the backburner, found out said friend died a few years back. Confronted the voice on the phone, and they revealed to me they were an AI program.
This is just me projecting, but we're at the "Novelty" phase right now. Like how the Automobile and Motorcycle were viewed when they first came out.
granted, this is a dream, but how real it was caused me to snap back into reality.
Imagine this: AI becomes so advanced, that we'll have something akin to VEGA from DOOM 2016 and Eternal. Silly, I know, but think about it. Within the next decade, maybe even sooner, we'll have AI programs capable of using voice that are so far advanced, the line between real and fake will literally become nonexistent bar you're involved from development of said program.
To summarize my dream in a few short sentences, it's like a twilight zone episode. Friend called me up whom I haven't talked to since high school. Did some catching up. Wanted to hang out, but kept getting put on the backburner, found out said friend died a few years back. Confronted the voice on the phone, and they revealed to me they were an AI program.
This is just me projecting, but we're at the "Novelty" phase right now. Like how the Automobile and Motorcycle were viewed when they first came out.
- 7,573
- Horbury, West Yorkshire
- GTP_Sprite
If they can eventually create a looping AI system as suggested here, then that’s where things could get messy?
arstechnica.com
Hype grows over “autonomous” AI agents that loop GPT-4 outputs
AutoGPT and BabyAGI run GPT AI agents to complete complex tasks iteratively.
- 7,573
- Horbury, West Yorkshire
- GTP_Sprite
You should read Silver Screen by Justina Robson.I don't really delve in messing around with AI programs like chatGPT, but I have been thinking about how advanced AI is getting.
granted, this is a dream, but how real it was caused me to snap back into reality.
Imagine this: AI becomes so advanced, that we'll have something akin to VEGA from DOOM 2016 and Eternal. Silly, I know, but think about it. Within the next decade, maybe even sooner, we'll have AI programs capable of using voice that are so far advanced, the line between real and fake will literally become nonexistent bar you're involved from development of said program.
To summarize my dream in a few short sentences, it's like a twilight zone episode. Friend called me up whom I haven't talked to since high school. Did some catching up. Wanted to hang out, but kept getting put on the backburner, found out said friend died a few years back. Confronted the voice on the phone, and they revealed to me they were an AI program.
This is just me projecting, but we're at the "Novelty" phase right now. Like how the Automobile and Motorcycle were viewed when they first came out.
It toys with the idea of a Ghost/Human conscience in AI form being part of a bigger AI system as such.
Edit
Sorry double post, thought I’d quoted it in the post above. I’m up past my normal bed time.
Last edited:
- 1,274
- ^mfw that Midget II
- Silhouette Drops
- 16,483
- Grea'er Laandan
- UKMikeyA
- UKMikeyA
Kinda reminds me of when Google's Cadie project switched itself off and went full Roy Batty.That's totally going to be the next generations Voight-Kampff isn't it?
CADIE's personal World Wide Website
web.archive.org
Last edited:
- 1,274
- ^mfw that Midget II
- Silhouette Drops
That last paragraph is something else.Kinda reminds me of when Google's Cadie project switched itself off and went full Roy Batty.
CADIE's personal World Wide Website
- 2,156
- United Kingdom
Photographer admits prize-winning image was AI-generated
German artist Boris Eldagsen says entry to Sony world photography awards was designed to provoke debate
- 28,410
- Glasgow
- GTP_Mars
Schumacher family planning legal action over AI ‘interview’ with F1 great
Michael Schumacher’s family are planning legal action against a German weekly magazine over an ‘interview’ that was generated by artificial intelligence
When artificial intellegence meets human crass stupidity...
- 4,819
- Kentucky
- barrett5
- BarrettMopar
TruthGPT seems more like a cryptocurrency scam than useful AI but it might generate enough controversy to at least get more people noticing AI and finding AI that can help.
truthgpt.one
TruthGPT
THE TRUTHGPT CHAT BOT IS GEARED TOWARDS ENLIGHTENING THE WORLD, GIVING UNBIASED ANSWERS TO ANY QUESTION YOU CAN THINK OF.
truthgpt.one
Last edited:
- 2,156
- United Kingdom
So did that song really use AI for the voices?
www.theverge.com
What’s really going on with “Ghostwriter” and the AI Drake song?
Something doesn’t make sense.
www.theverge.com
Similar threads
