The Week in AI: AGI In a Few Years? (AI Mind Reading, Bing Upgrades, MJ5.1, New Open Source LLMs)
The Week in AI: AGI In a Few Years? (AI Mind Reading, Bing Upgrades, MJ5.1, New Open Source LLMs)
The Week in AI: AGI In a Few Years? (AI Mind Reading, Bing Upgrades, MJ5.1, New Open Source LLMs)
In some sense, we’re all doing AGI’s bidding already.
Large Language Models are feeding on our data. Our usage of LLMs goes a long way toward fine-tuning and providing feedback. Companies are racing toward constructing a dwelling place for AGI to inhabit and all the while, you and I are animating its triumphant march toward the temple.
The betting markets are beating the drums at a steadily increasing tempo:
And gamblers aren’t the only ones expecting it to happen soon:
The best way to predict the future is to create it. If AGI is ever to become a thing, having millions of people willingly or unwittingly contribute their intelligence and be the sentience of it, what we’re seeing now is what it could look like—kind of like a neural network.
Perhaps this is one reason the “Godfather of AI” stepped away from Google and is now issuing dire warnings.
Neural networks that learned from vast amounts of digital text were a breakthrough in a long chain of major and minor steps toward where we are. Dr. Hinton thought these networks (and LLMs) were inferior to how humans handled language—until he changed his view.
“Maybe what is going on in these systems is actually a lot better than what is going on in the brain.”
Run the chain on that thought and:
“Look at how it was five years ago and how it is now. Take the difference and propagate it forwards. That’s scary.”
This, along with other warnings and bad omens from all over the place, and we find ourselves watching as Congress, the White House, and CEOs of tech companies with a vested interest in protecting their turf descend on Washington:
They’re not meeting out of the goodness of their hearts and because they have your best interest (or rights) in mind. The past few years have made it abundantly clear that they do not care about your rights and will suspend any laws to achieve their aims.
They all know the future of AI is being determined now. The goal is regulatory capture, formalizing monopolies, concentrating AI power in the hands of a few companies, and centralizing AI command together with D.C.
You can expect to see legislation that favors the few, keeps AI out of your hands as much as possible, restricts whatever use you’re allowed to have to a minimum, and leverages the power of AI to surveil, control, and dominate your life—with the full force of tech companies, Silicon Valley, and the Federal Government behind it.
You’ll be fully immersed in it—not for your benefit, but for theirs.
And by fully immersed, I do mean you’ve got no choice because AI can read minds now:
Get more info on this website, and you can read the paper here.
And you can stock up on tinfoil hats here.
With all this in mind, perhaps we’ll see something like this play out soon:
Perhaps another contributing factor to their desire to lock down AI is also driven by an alleged leaked memo from Google: "We Have No Moat, And Neither Does OpenAI".
Who knows how legit this “leaked memo” is, but the concerns raised seem mostly correct. It is true that LLMs on a phone and scalable personal AI are a threat to rising AI Regime.
The small opening to an escape might be the rapid rise of open-source LLMs and AI running on your own devices.
Perhaps there will be other companies arming The Resistance, too:
Related to that, your own personal AI is becoming a thing.
In other AI news around the internet
Let’s finish this week on a more positive note with a quick roundup.
Microsoft upgrades Bing Chat with new superpowers, including image and video answers, plugins, chat history, and a better Edge integration. They’re also removing the waitlist and you can now get instant access (with a Microsoft account).
OpenAI releases Shap-E, a 3D generative modeling tool. Get models made from text.
Midjourney 5.1 is out and text-to-image is reaching near photo-realistic in 2023. Improvements include greater accuracy, fewer unwanted borders or text artifacts in images, and improved sharpness. Peruse a thread like this and be dazzled.
ChatGPT is your new data scientist team member. Have data analyzed and turned into visuals:
Get an in-depth breakdown of what the Code Interpreter and more can do in this piece from Ethan Mollick.
The Week in AI: AGI In a Few Years? (AI Mind Reading, Bing Upgrades, MJ5.1, New Open Source LLMs)
 | A guest post by
|