Open AI 2

(cont'd from yesterday's post)

Open AI wrote the "chatbot" (ChatGPT) that some students are using. It was not designed for cheating, but to perform a task which you give to it in natural language (not code). A WSJ writer went back to an advanced placement English class to test it. She gave her assignment to the AI program, and it wrote an essay which got the teacher's approval.

When this AI tool launched last November, it had a million users within five days.

Generative models coming from Open AI can generate music and images as well as text, with spoken natural language prompts. The image below was generated from the prompt, "Teddy bears working on new AI research underwater with 1990s technology" 

Potential ethics problems abound. Deceit? Already happening. At this point, you won't know if a person or an algorithm (written by a person) wrote an article you're reading. Bias and manipulation? Sure, just write the algorithm to exclude any idea you don't agree with - like the social media companies do.

Elon's view is that people with good motives must be involved in the inevitable development of AI to move it in the right direction. He wants Open AI's research to be open-sourced, which means accessible, not secret.

Open AI 3

Follow up to these posts

Though being one of the founders of Open AI, Elon Musk resigned from its board of directors some time ago and is not associated with it anymore. He intended that it would be non-profit and open sourced. 

photo

Today he says it is profit-driven, controlled by Microsoft, and trained to be "woke" -- its algorithms give privilege to politically left people and ideas while censoring those that are not left-leaning, something that social media does routinely. Open AI executives agree that it's biased, and say that bias is very hard to change.

So Elon is recruiting artificial intelligence researchers to work on a new system that will compete with ChatGPT but will not be woke.

from Deseret

Open AI 1

Open AI does research in artificial intelligence (AI). There's a for-profit wing which develops new commercial applications for sale, and a non-profit side with no commercial sales function. Elon Musk was one of the original founders back in December of 2015, but he's no longer on the board.

It reflects Elon's well-known fear that artificial general intelligence will eventually threaten human beings. A simple statement of their mission from the website: "to ensure that artificial general intelligence benefits all of humanity."

(cont'd tomorrow)

"I" in AI? #1

 All this progress in artificial intelligence makes us think more deeply about what we mean by "intelligence." As you already know, AI is used all over today's culture - including diverse applications like farm management, your kitchen appliances, business and military virtual reality training, health care.

image

It looks like real intelligence. But it's still "narrow" AI which simply performs a task it is commanded to do based on the data it collects and on the goal that it is designed to reach. 

Last November, ChatGPT was brought to the public by its creator Open AI. It generates text when given a command in ordinary language, and that makes it look like it has "general" AI (the human kind). But it doesn't. It is still just computing data as its programming dictates. 

It doesn't know what the words mean. It doesn't know or care whether its sources are true. When asked to reveal where it got the information it bases its claims on, it just makes them up. 

from Mind Matters

(cont'd tomorrow - ChatGPT  errors)

AI: wrongful death

Are AI platforms or their owning companies responsible if they produce terrible outcomes? The Starbuck suit (yesterday's post) against Google may help define this question. But it's not the only lawsuit trying to hold AI accountable.

A 16-year-old boy committed suicide last year, and his family is suing Open AI for wrongful death.

Never happened 1

Artificial intelligence has advanced greatly since Open AI's ChatGPT, which was itself a breakthrough advance in 2022.

No business wants to be left in the dust, so staff work diligently to find every possible way to use the new technology. More generative large language models (LLM's like ChatGPT) are being developed by tech companies. Gemini by Google is one of them.

Shortly after its launch, Gemini triggered alarm. When asked by users to generate pictures of historical figures, it showed them . . but not as they were. Minority races were substituted for white people and dressed in the style of their century. 

People are learning, if they didn't know it before, that AI does not necessarily report true and factual information. 

Gemini and all other artificial intelligence only follows the instructions (algorithms) of its creators. It does not "think things through" or look for truth. "AI can only learn from what it’s been given" by its developers--who have their own biases.

from "Google Gemini Presents a Pattern That Never Happened"

(cont'd tomorrow)

AI merge

(cont'd from yesterday's post)

According to Elon Musk, the average person won't begin to consider AI a threat to humans until he or she sees "killer robots walking down the street." But in his opinion, that will be far too late. 



A prime example of AI: 

Deep Mind is Google's on-going project to create "digital super intelligence," "vastly smarter than any human on earth and ultimately smarter than all humans on earth combined." It has "administrative access to Google's servers . . it could take complete control of the whole Google system which means it could do anything . . " including interact with all your data. 

His recommendation? "If you can't beat 'em, join 'em . .  It's incredibly important that AI not be other. It must be us . . I'm afraid we're either going to have to merge with AI or be left behind."

He admits that he could be wrong about this and says he's open to anyone coming up with a solution that's better.

Job destroyer?

Fear of losing their jobs to AI is a real worry for about 40% of workers. It's true that artificial intelligence replaces some tedious, repetitive jobs, but it can even do something like a writing job. 

In fact, that's a key point of the SAG-AFTRA strike now going on. Faces and voices are easily replicated, while something like ChatGPT can write realistically (but not creatively). Actors and writers want regulations to protect them.

photo

But this think-tank writer says there are also opportunities coming: "Perhaps . . AI will turn out to be a tide that lifts all workers rather than a job-destroying technological tsunami."

For example, a study of customer service at call centers found that AI helped all to be more productive, and lower-skilled workers quickly grew to the level of higher-performing workers. In general, it may support all workers in their areas of weakness.

As usual, innovation can open new opportunities for those who prepare for it.

from AEI

(cont'd tomorrow: AI at Amazon)

Fear of AI 4

(cont'd from yesterday's post)

While researchers are building their own artificial intelligence to race against competitors, users are thinking of ways to use it. Khan Academy, for example, is promoting "Khanmigo," their "artificially intelligent personal tutor" that can talk a student through a problem like a live, personal tutor would. It sounds impressive.

Massachusetts Institute of Technology (MIT) says, "ChatGPT is about to revolutionize the economy. We need to decide what that looks like." Since ChatGPT came out last November, businesses are intent on developing it for their benefit. Nobody wants to be left behind.

We've seen the open letter from tech experts, we've seen Yudkowsky's advice to shut it down, now there's a third big warning about AI. Just a few days ago, the "godfather of AI" was interviewed in the NYT:  he regrets the work he devoted his life to because of how A.I. could be misused. Wow.

(cont'd tomorrow)

Fake sources

Follow up to this post

So, Open AI produced ChatGPT and it's being used to write text, like an essay (it can also generate art and music). You can tell it in your own words what to write, then it will survey some internet sources and come up with an article. 

image

This author has been using it and noticed something interesting. He asked it to write a sample blog post and to include its online sources, where it got the information it used. It did give him sources, all right - fake, made-up websites and articles.

He tested it again, and it gave him a list of five links to articles at various URL's. They were all fake. He told it, "None of those articles actually exist." It apologized and gave him another list of fake links. 

Actually, I generate text too. Every weekday I publish a post I've written, and . . and . . give you the actual link to the actual article or book I based my post on. You can always check my sources because the links are real.

You're welcome.

from Mind Matters

Big Brother 2

(cont'd from yesterday's post)

China is quite open about surveillance. The government monitors almost everything people do on the internet or their phones, and local units are putting up hundreds of millions of cameras that use facial recognition artificial intelligence (AI).

Now totalitarian rulers have a new tool to use: "Data and social scientists from the University of Chicago have developed a new algorithm that forecasts crime . . ." Breakthrough! "The model can predict future crimes one week in advance with about 90% accuracy." How convenient.

Similarly, "Scientists at China’s Comprehensive National Science Center in Hefei claim to have developed mind-reading artificial intelligence that measures the loyalty of Chinese Communist Party members."

image

Mind-reading algorithms? Crime predicting algorithms? What opportunity for abuse. "They will surely be used to falsely imprison political enemies . . to blackmail people for real or imagined misdeeds."

from Mind Matters

COSM 2023

ChatGPT, launched less than a year ago, is now the fastest-growing app of all time. After only two months, 13 million customers per day were using it last January.

Duolingo (learn a language), Expedia (travel planning), Coca Cola are just a few businesses who already use it (or its improved versions) to give their customers better, more personalized service. 

AI has certainly reached a new level. It's everywhere and it's surging. Many are alarmed. Some say it must be slowed down or even shut down entirely until cooler heads can figure out how to regulate it.

At the very least, civil discussion among knowledgeable people with different viewpoints needs to happen. I know of one event that's organized to do that very thing: COSM 2023 in Seattle.

"COSM is an exclusive national summit on the technologies remaking the world as we know it. The mission of the conference is to stimulate debate and deliberation amongst industry leaders . . ." It's open to the public.

from Mind Matters

Stop advancing 1

Artificial intelligence inspires hope. Hope drives technology companies and their researchers to make ever-more-intelligent programs to manage enormous amounts of data that improve every single sector of our economy and help us achieve great new things for the common good.

But that's not the only hope that drives technology. There's also the drive for profit and for control -  neither of which is evil in itself, but can easily be used for selfish or evil goals. Every technology breakthrough for the common good can be corrupted, like what has happened to social media.

Some technology leaders, including Elon Musk and Apple cofounder Steve Wozniak, are really worried.

They got together to write an open letter: It calls on AI labs to quit advancing for six months due to "profound risks." 

photo

(cont'd tomorrow)