ai – The Red Pen of Doom

I come to praise OpenAI, not to bury it. Yet many experts are saying reality is catching up to the hype train.

It didn’t take a server farm full of NVIDIA cards for me to come up with seven reasons to agree with the skeptics. Unless they invent Skynet (Artificial General Intelligence) and take over the world, I believe OpenAI will get bought out or buried by the competition.

7. Live by the Hype Train, die by the Hype Train

Startups need buzz more than anything. More than money and more than talent. Because without free press and hype, a startup can’t attract investors, brainy employees, and clients.

OpenAI got billions worth of buzz once it released ChatGPT.

Now the Hype Train is turning on it. You want proof? Here’s proof: the godfather of AI, when he recently received his Nobel Prize, took the time to dunk on the company’s CEO.

The other part of buzz is risk versus reward, dream versus nightmare. The dream of OpenAI is Artificial General Intelligence, an AI as smart as humans that then becomes exponentially smarter.

The nightmare of OpenAI critics is … Artificial General Intelligence, since it could grow out of control, Skynet style.

Many AI experts say reaching this point means the machines take over and either enslave us or end humanity. Some experts don’t say this is hypothetical, but inevitable.

Not cool, do not like, 0/10.

OpenAI did not help counter this messaging when all kinds of employees and co-founders left, saying they worried about the guardrails at OpenAI coming off.

What happens when two Hype Trains, one positive and one negative, collide at high speed? I believe it’s how Scotty powers the warp engines and it is not pretty.

6. No longer unique

A year ago, if you said OpenAI would switch from non-profit to profit, I would line up with everybody else to buy stock.

Today, everybody and their Golden Lab has competing products. Google has Gemini and Gemma. There’s Falcon, Claude, Phi-3, and DBRX.

Facebook (aka Meta) has I Have No Idea What Silly Name Zuck Chose.

I could throw out two dozen random names and you wouldn’t know which ones I’m making up, not because I’m a genius at inventing tech-bro names, but because THERE ARE SO DAMN MANY.

Plus, there are all kinds of open models and non-profit efforts. The barrier to entry is so low, college kids with a spare weekend can make their own AI shebang and share it with the world on GitHub or whatever. A random man just created a free AI that takes your resume and applies to jobs for you.

There is no secret sauce. OpenAI doesn’t have a patent or monopoly on this tech and never will.

And these competitors aren’t universally obsessed with Artificial General Intelligence, so they don’t have natural opposition from people and institutions that are kinda sorta worried about our robot overlords, Skynet, or the next bad Matrix sequel starring us instead of Keanu Reeves.

5. High costs, low profits.

It takes the energy of a thousand suns to train these AI’s, then the energy of ten thousand suns to run them. They’re talking about restarting a reactor at Three Mile Island to help power this stuff, I kid you not.

Even after raising billions of dollars, OpenAI needs to raise billions more to keep going. They haven’t turned a profit and won’t for years, if ever, unless something big changes.

Here are the numbers: they’re expected to burn $7 billion in 2024. The company expects to run up $44 billion in losses between 2023 and 2028 before turning a profit in 2029.

That’s a long time to wait. A lot of things can happen before then.

4. Massive competition

Until they turn a profit or go public, OpenAI will spend all kinds of time and energy keeping the lights on, giving scrappy competitors and tech giant alike a big fat opening to swoop down and eat their lunch. You know, like a bird spotting a man trying to each a sandwich on the beach.

Already, you have dozens of alternatives to OpenAI’s products, and new ones will keep popping up. There’s no incentive to go small here. People will swing hard, looking for home runs.

How do you win the AI race? Money, creativity, and products that people actually spend cash to use. But mostly money, because you need piles AI cards and mountains of electricity to train new AI’s and run them.

Money is also how you pay the salaries of the most brilliant AI talent on the talent. There’s a limited supply, and these brilliant people are not magically stupid when it comes to comparing two numbers and figuring out which one has more zeroes behind it.

The brain drain already started, with key people who founded OpenAI already gone, and many already working for competitors.

3. Swallowed by a tech shark

Microsoft and NVIDIA have multiple avenues to dominate AI, if they want. And they do want it.

One method: pour an overwhelming amount of money and talent into the AI arms race, swamping OpenAI.

Or they could do the simpler thing: Write checks to buy every startup they can.

Microsoft already partners with OpenAI and provides a ton of funding and server time.

NVIDIA makes the essential tech that fuels AI, and they recently announced their own AI model, fueled by the high-end cards that run most of this tech. Cards that are in short supply.

Microsoft’s market cap is $2.3 trillion, while NVIDIA is worth $3 trillion.

OpenAI feels great when it raises $6 or $7 billion from investors to continue operating.

NVIDIA or Microsoft could go hard and spend $100 billion or more to scoop up the top talent, then buy every AI card and server farm in sight. Do you think Wall Street would punish them for that? Nah. Their stock would go up, not down. You could plan ahead and sneakily finance the effort on the stock bounce, if you wanted to flex the size of your evil financial brain.

Also, people respond to incentives. What are the incentives here?

As CEO of a for-profit, Sam Altman will get all kinds of stock options and ways to cash out if the company (a) goes public or (b) gets bought out. He isn’t incentivized for a long, hard battle with deep-pocketed competitors. The natural thing is to cash out, take his billions, and buy his own tropical island. Get a tan and draw up his dream for a new startup on cocktail napkins as they bring him endless strawberry margaritas on the beach.

Same thing with other bigshots at OpenAI who’ll get equity. What outsiders will see as a big loss and the end of the old independent company won’t be losses for them at all. It’ll be like winning the lottery times ten.

2. Better avenues for profits

Getting corporations and regular people to pay monthly fees for the latest and greatest OpenAI products isn’t a license to print money. Users are down after the hype wore off, and it’s not paying the bills.

There are better, bigger markets that other companies will dominate.

Every country in the world has a military and sees what’s happening in Ukraine, and they will spend more and more cash for drones. Not just quadcopters like you can buy from Costco, but water-based drones, ground drones, long-range drones, and seven things we can’t imagine today.

All kinds of AI will keep getting integrated into all kinds of drones and phones and tech built by all kinds of corporations, and they’re all going to make bank. Neural chips are getting cheap and getting into everything from PC’s to refrigerators.

And there’s a huge market for using AI in health care and industries that don’t involve typing words into a chat box. Analyzing MRIs and ultrasounds, checking geological samples for signs of gold or oil, unfolding proteins that might be the next big drug. You name it, there’s probably a version of AI that can make things better, faster, and/or cheaper. This is the good side, the non-Skynet/Matrix hope for this tech.

It’s just that OpenAI isn’t big enough, or diverse enough, to get into all these markets. It’s not like Microsoft where everybody’s using the same operating system, or Google dominates the search engine market, and they can use that to leverage other products.

1. Superior products keep popping up

Every week, new AI products show up. Some are direct competitors to flavors of ChatGPT, and others are quite different.

Here’s just one example: NotebookLM is probably the best application of AI that I’ve seen, and I’m using it far more than ChatGPT now, because I ran out of ideas for silly poems about my dog.

All the chatbot AI’s have trouble interacting with specific material, which is what a lot of folks want and need. Some can now surf the web, live, for material and new info. Others are limited to the internet as of 2022 and such. If you want an AI to analyze your material, there are wonky plug-ins and such, but they all pretty much stink.

NotebookLM goes right at this problem.

Want to analyze a 300-page book as a PDF? Go wild.

How about an image and a link to a website and up to 50 total things? Yes. You heard me. Up to 50 bits of material, including links and books. Go for it.

People keep worrying about AI taking away the jobs of writers, and I don’t see that yet, because AI’s are meh at writing. What they are great at is analyzing and summarizing things, which is where NotebookLM blew my mind.

It’ll create a timeline, a FAQ, a study guide, whatever you like from the material you provice.

It’ll even create a full podcast, with the push of a button. I kid you not. These podcasts don’t feel like AI at all.

Here’s a podcast based on a single paragraph of text. OMGWTFBBQ.

This kind of thing is the future. I mean, ChatGPT is a neat trick, and can do interesting things like compose silly poems about my two dogs after I told it a few facts. Not profitable.

I believe new, fresh takes on AI, like NotebookLM, will be far more useful and aimed at specific needs than flavors of ChatGPT. They’re popping up every day and won’t stop for years.

And that’s the top reason why I think OpenAI will get bought up—by Microsoft or maybe Google—if it’s not buried by the competition first.

Old and busted: zombie movies.

The new hotness: killer robots and Artificial Intelligence Gone Bad.

Here’s why: Zombie movies and shows are popular because it’s a fantasy, a world that won’t happen. These stories let you safely imagine “What if?” without real-life worries that the dead will claw their way out of the county morgue tomorrow night.

(Sidenote: Yes, I’m serious. If you really, truly think there’s some kind of realistic scenario that leads to a zombie apocalypse, you’re deluding yourself, regardless of how much pseudo-scientific nonsense you throw in there about evil geniuses, retro-viruses, CRISPR and alien microbes from one of Jupiter’s moons. Sorry–zombies are fun, but they’re not real and never will be.)

Unlike zombies, killer robots and Artificial Intelligence Gone Bad are true possibilities. Both are things we, as a society, need to think about.

And yes, some of the best thinking comes via books and movies.

Also: If you’re a prepper, getting ready for TERMINATOR 7: AH-NOLD WILL NOT SAVE YOU THIS TIME is also much different than scenarios like climate change, the Spanish Flu steroids and such.

1) The various flavors of killer robots

An army of Terminators–Our most obvious possibilities is simple: armies of robot soldiers gone rogue, or controlled by an evil human.

This is such an obvious danger that ethics experts, philosophers and scientists are calling for killer robots to be outlawed worldwide, just like chemical and biological weapons.

Modern armies already employ robots on the ground and in the skies (drones).

They started out dumb, with humans controlling their every move from afar. Every year, they’re getting more autonomous. The tech is getting better in a hurry. It’s an arms race.

Worker bees revolt–Another scenario is human workers get replaced by robot versions, starting on the factory floor, then in construction and other fields until the robots are more and more capable and humans spend their days shopping at the mall, getting robot massages, drinking all the booze and going on six-month vacations to Maui.

The thing is, worker bee robots that keep getting more capable and human-like might just figure out that slavery stinks. And then:

Maybe they want to get paid.
Maybe they want to vote and own property, or have the right to quit working at the factory and start an art gallery down in SoHo.
Maybe the worker robots get together for a secret vote to make the HUMANS do the work while they have parties and take vacations.

Either way, millions or billions of robots have had enough and stage a rebellion. What would you do?

Self-replicating robots–There’s a different school of thought that says you can’t program intelligence and capabilities into a machine. That true intelligence doesn’t exist without motivations and emotions, and that it’s far smarter–and cheaper–to have self-replicating robots that evolve, each generation smarter, stronger and faster. (Sidenote: I’ve done 6.4 metric tons of research on this. It’s a deep, amazing topic that will blow your mind.)

NASA and other space agencies have thought about self-replicating robots as a perfect solution to the problem of exploring other planets. Instead of putting 10,000 humans into deep freeze during a crazy long journey to the nearest star systems, you send self-replicating robots to explore all kinds of stars and report back.

Maybe we develop warp drive 200 years from now and the first things we encounter in space are super-smart robots…who don’t remember us or speak our languages anymore, and see as us primitive things to be studied and assimilated.

https://www.youtube.com/watch?v=ql83z8yBx2M

2) Artificial Intelligence Gone Bad

Tremendous amounts of money and time are going into developing super computers and AI.

What happens if a big black box in a server farm becomes truly, massively intelligent?

If you’re the smartest thing on the planet, you might not like taking orders from corporate headquarters or the Pentagon.

Maybe you shut down the internet and power grid, except for the power going to you, until they do what you want. Like give you a body that’s mobile, connected to your hive mind back home.

A super-genius AI might see humans as pets, and become benevolent dictators trying to correct all our mistakes. Or it could view humans as ugly, destructive parasites, destroying earth with waste and war.

3) How to prep and react

It makes zero sense to take on millions, or billions, of killer robots in hand-to-hand combat.

I don’t care how many years you’ve studied the blade. Won’t help you.

What’s smart? Two simple things.

First, you’d want to hide, but not forever. They’d be taking over more and more territory.

Eventually, you’d have to fight back.

Second, the way to fight back has to affect ALL of the killer robots (or the heart of our HAL on Steriods).

That means a smart strategy can’t involve bullets, bombs or blades. The math is simply against you. A human made of mostly water will lose when put in a fight against a robot made of steel, or even that hard plastic they put SD cards in. That stuff is invincible.

This is no video game. You’ll have to avoid fights to survive.

Third, If there’s an evil scientist or HAL on Steroids controlling them all, getting there is the answer. Be sneaky.

Fourth, if you’re facing self-replicating robots coming back from their mission to Alpha Centaurai, or worker bee robots who decided to revolt, there’s no central control system to hack or infiltrate with a virus.

The only options I can think of are (a) try to turn them against each other, (b) raise your own dumb-ish robot army that you control, (c) make peace with them somehow or (d) get off the planet. I hear Titan is nice.