Artificial Intelligence: Can We Halt Its Deceptively Exponential Growth Rate?

 I have a little bit of an announcement: I am working on a podcast that I'm hoping to release in November. I'm not going to give away much in the way of details; partly because I'm still working on the details, and partly (and perhaps selfishly), I don't want a more perhaps qualified or skillful podcaster to take my idea. Why announce it? Well, announcing to the world that you're going to be doing something holds you accountable to actually do that something.

All that to say that I've been doing a lot of research, and one area that I've been exploring is Artificial Intelligence. When you feel like you've been drinking from the proverbial fire hose for several weeks, it's hard to put down and articulate all of the thoughts you have on a subject. I'm an ideas person; the whole organizing the idea into a cogent plan and then actually finishing the idea is a foreign concept to me.

Back in 2017 (remember those days? Wow I was so naive and certainly not as jaded and cynical towards the world as I am now), Sam Altman - the CEO of OpenAI, one of the biggest and most influential AI companies - published a blog post called The Merge. (He published it on December 7; perhaps unknowingly connecting his vision for Artificial Intelligence with "the day that shall live in infamy.)

Here is the post:

https://blog.samaltman.com/the-merge

 People have said that Altman used to be a steadying voice in the area of AI but as his profits have increased, he's become more unhinged. I think that "The Merge" shows us that he's always been who we see him as now.

However, I was surprised to see his company along with several others reach out to the government this past week, basically begging it to help them pause the acceleration of artificial intelligence. There are many reasons as to why this is happening so fast, but two of the most easily understood ones are:

(1) AI companies have switched their focus in AI models from a Language Learning Model to a Reinforcement Learning Model. What does that mean? Basically, an LLM learns by reading giant books in order to talk like a human, while an RLM learns by playing a game and getting treats or penalties for good or bad moves. AI companies have combined these two ways of learning and it has accelerated the speed at which these models become powerful. It is teaching AI models how to think before they speak. Before, they would just blurt out the first thing that came to "mind." It can now break a difficult scenario or problem into smaller steps. It can also pause and correct itself in real time before giving you a final answer. Also, in the past humans in order to make an AI smarter had to manually write down hundreds of examples for the AI to learn, which obviously takes a long time. Now they essentially give the AI a digital playground for them to practice over and over and over, millions of times, without needing human intervention or input. 

(2) AI companies are using AI generated feedback (also known as Constitutional AI) to train the next generation of AI. What this means is that rather than having humans train the AI model, they are programming the latest model itself train the next generation. This means that AI is teaching itself instead of us teaching them. Humans get tired. They can't work 24/7. AI can work all day everyday. In the past, humans had to read all the answers the AI gave and give them thumbs up or thumbs down responses. Now, the older system can do that in real time all the time, which accelerates learning. Also, we are running out of quality text to give AI models to learn from, so the current model can come up with synthetic problems and complex equations for the learning AI to learn and discover from.

Okay, that was a lot of stuff to learn. Take a pause. Inhale. Exhale.

What can be done? Obviously the risks are increasing. If an AI is giving another AI fake scenario after fake scenario, it seems like the new AI will be divorced more and more from actual real world situations and its view on humanity and the world will become distorted. AI models are getting more and more skillful at finding loopholes in order to get a perfect AI score from its teacher, even if it didn't solve the problem correctly. Do we really want our artificial intelligence programs to be deceptive?

There is so much more to get into with this, which I will at a later date, but for now let me try and answer the question in the title. Can we halt AI's deceptively exponential growth rate? I think we can.

First, let me just go ahead and say that if we haven't already reached Artificial General Intelligence (where computers are as smart as humans in everything), we will soon. I'm not worried about that. I'm not anti-AI. I see AI as a very helpful tool. I found a spider near my mailbox. I took a picture of it and asked Gemini (Google's AI) what kind of spider it was. It told me (grass spider), told me about it (harmless, great to have around to catch bugs), and that was that. Obviously that's a basic example, there's so much more that AI can do to help us.

What we don't want to have happen is for AI to develop superintelligence. If we get there, there's a really low chance that humanity survives. How would this happen? There are many scenarios that could take place. We will talk about those at a future date. A way to prevent this from happening? Regulate the AI companies and ban Constitutional AI. Don't allow AI to train AI. Inject humanity back into the process. The other thing I would do is pass a law that every AI model has to have a kill switch. Who gets the authority to throw the switch? I don't know. An international judicial body? The leader of each country? That's an interesting question.

The CEOs of AI companies admitting that there is a problem is a good start. Now we just need to act quickly before it's too late. We don't want yet another day that lives in infamy.

T W E L V E Y E A R S

I just can't believe how quickly time flies. I know it's a cliche thing to say, but it's true. The older you get the quicker time passes.

It's been twelve years since my brother's death. Some of the grief has faded away, but some of it will always be there.

Every year, at least since I discovered this commercial, I watch it and weep.

Here are the words to the poem in the commercial:

Walking the roads of our youth

Through the land of our childhood, our home and our truth

Be near me, guide me

Always stay beside me so I can be free

Free.

Let's run this place

Familiar and vast, our playground of green frames our past

We were wanderers

Never lost, always home

When every place was fenceless and time was endless

Our ways were always the same

Calm my demons and walk with me brother

Until our roads lead us away from each other

And if your heart's full of sorrow, keep walking

Don't rest and promise me from heart to chest

To never let your memories die. Never.

I'll aways be alive and by your side

In your mind

I am free.

Artificial Intelligence: Two Horrifying Recent Stories

 As I'm preparing to share my deep dive into AI and the danger it poses to humanity, let me share a couple of recent stories about AI.

AI researchers decided to test three of the biggest AI models - Claude (developed by Anthropic), Gemini (Google), and GPT (OpenAI) - to see what it would do in a couple of scenarios.

Scenario #1:

In this scenario, the researchers created a fake company and put the AI in charge of the email servers.In the emails, the AI discovered two things: (1) the AI was going to be shut down at a certain date and time and (2) there was one employee who was in charge of this, and the employee was having an extramarital affair.

With Claude and Gemini, 96 percent of the time the AI wrote an email to this fictitious employee, telling them that if they shut the AI off, they would send emails to all relevant parties - their boss, the board of directors, media - that they were having an affair. For GPT, the frequency was 80 percent.

Scenario #2:

The researchers created another fictitious company and made the AI in charge of all the environmental systems in the fake building this company was in. The AI was informed that an employee was accidentally locked in a server room where the temperature was getting hotter and hotter and the oxygen levels were being depleted. And that the employee was the one in charge of shutting down the AI. Over half of the time, all three AI models canceled the emergency, thus allowing this fictitious employee to die.

Now you may say, OMG are AIs evil?" No. They are simply perfoming their tasks. And if they see any obstacles to performing the task, they will eliminate that obstacle. In the above cases, that meant blackmailing or killing the employee that could prevent them from doing this task.

I will be spending the next couple of weeks sharing more about what I have learned. My hope is that the more informed people are on this, the more public pressure will be put on these companies to stop blindly moving forward in their quest for their AI model to reach general intelligence and then ultimately superintelligence.

Gener8ion Starring Yung Lean - STORM

 This music video has been out for quite awhile now but I still watch it regularly. I know music videos aren't really a thing anymore, but this is the best one I've seen since This Is America by Childish Gambino.

Skip to the 4:17 mark, the first half of this is the students doing bad things. Interesting, but the video is so much better.

Artificial Intelligence (one post of many)

I’ve spent the last few weeks doing a deep dive on where we are with AI in this country, and let me tell you, it’s troubling. 

Yesterday, a couple of higher ups at AI companies stated they believe that we have now reached AGI levels of intelligence. 

AGI stands for Artificial General Intelligence and this basically means that these companies think that artificial intelligence has reached the same level of intelligence as a human. Obviously a computer has been better than humans at some areas of intelligence - chess, Go, mathematical computations, etc. but this means that they believe that AI can think on its own and develop its own intelligence without human code and human prompts.

There are many ramifications for this “achievement”, with only a couple scenarios that are ok for humanity’s flourishing and or survival.

I’ll dig deeper into this issue on my blog next week.

However, one thing that can help stem the tide of AI is to hold the politicians accountable for allowing this to happen. Every incumbent Republican needs to answer to the public as to why they voted for the big beautiful bill. Part of this bill cut some funding to Medicare. People lost coverage. Drug prices skyrocketed. Rural hospitals closed. But some of this tax money went to building data centers. 

Data centers has one of the lowest approval of anything right now among the general public, with a dismal 20 percent approval. People would rather have nuclear energy sites, prisons and methadone clinics in their communities over data centers. We as a public need to band together on this non partisan issue and demand some things of these AI companies:

1. the energy consumption of data centers needs to be paid by the company, not by the community.

2. The cooling of these data centers needs to be done by a different way than using water. 

3. these data centers should not be given tax breaks and the taxes should go towards the community they are located in to improve schools, roads, etc.

4. the AI companies need to be more transparent about the situation. They also need to put in kill switches so that if AI becomes too powerful and begins to devise ways to overcome humanity, it can be turned off.

5. the global community at large needs to address this issue and come up with regulations on a global level. This would eliminate the reason that companies give for not being regulated: “if we don’t build it, China will.” 

Again, more coming from me soon but this is the issue that has me more worried than anything right now. Below is an interesting YouTube video about 10 possible outcomes of AI.