AI Weekly: Is AI’s Alien Invasion Imminent?

Want aI Weekly free every Thursday in your inbox? Login here.

Is an AI alien invasion coming towards Earth? The VentureBeat editorial team marveled at the possibility this week, thanks to massive online traffic generated by a community article by Data Decision Makers with the unmissable title, “Prepare for Arrival: Tech Pioneer Warns of Alien Invasion.”

Written by Unanimous AI founder Louis Rosenberg, the column was certainly spurred not only by its SEO-friendly title but also by its breathless opening: “An alien species is on its way to planet Earth, and we have no reason to assuming she will be kind. Some experts predict it will be here within 30 years, while others insist it will arrive much sooner. No one knows what it will look like, but it will share two key traits with us humans – it will be intelligent and confident.”

But a more complete read reveals Rosenberg’s focus on some of today’s hottest AI debates, including the potential for AGI in our lives and why organizations need to prepare with AI ethics: “…while there are serious efforts in the AI ​​community, on to push safe technologies, there is also a lack of urgency. That’s because too many of us mistakenly believe that a sentient AI created by mankind will somehow be a branch of the human tree, like a digital descendant that shares a very human core. That is wishful thinking… the time to prepare is now.”

DeepMind says game over for AGI

Coincidentally, the past week has been filled with claims, counterclaims, and criticism of claims surrounding the potential for AGI to materialize in the near future.

Last Friday, Nando De Freitas, a senior researcher at Google’s AI division DeepMind, tweeted that “Game’s Over!” in the decades-long search for AGI after DeepMind unveiled its new Gato AI, capable of handling complex tasks ranging from stacking blocks to writing poetry.

According to De Freitas, the Gato AI simply needs to be scaled up to create an AI that can compete with human intelligence. Or as he continued to write Twitter, “It’s all about scaling now! It’s about making these models bigger, more secure, more computationally efficient, faster at sampling, smarter storage, more modalities, innovative data, online/offline… Solving these challenges is what AGI will deliver.”

Pushback to AGI and scaling

Many experts dismiss the claims of De Freitas and those of others who insist AGI or an equivalent is on hand.

Yann LeCun, the French computer scientist and senior AI scientist at Meta, had this to say (on Facebook, of course):

“Regarding the heated debate about the significance of recent advances in AI, it may be useful to (re)state some obvious facts:

(0) AGI does not exist. Reaching “Human Level AI” might be a useful goal, but humans are specialized too.

(1) The research community is making some progress towards HLAI

(2) Scaling helps. It is necessary but not sufficient because….

(3) we are still missing some basic concepts

(4) some of these new concepts may be “around the corner” (e.g. generalized self-supervised learning)

(5) but we do not know how many such new concepts are needed. We only see the most obvious.

(6) Therefore, we cannot predict how long it will take to achieve HLAI.”

Current AGI efforts as “alternative intelligence”

Meanwhile, Gary Marcus, founder of Robust.AI and author of Rebooting AI, contributed to the debate on his new substack, with his first post dedicated to discussing current efforts to develop AGI (including Gato AI), which he calls “old intelligence”. :

“Right now, the dominant workstream within Alt Intelligence is the idea of ​​scaling. The notion that the larger the system, the closer we come to true intelligence, maybe even consciousness.

The study of alternative intelligence is nothing new in itself, but the hubris surrounding it is. I’ve been seeing signs for some time that current AI superstars, and indeed much of the entire field of AI, are treating human cognition, ignoring and even ridiculing scholars in fields like linguistics, cognitive psychology, anthropology. and philosophy.

But this morning I woke up to a new reification, a Twitter thread that loudly expresses the Alt Intelligence credo of Nando de Freitas, a brilliant senior executive at DeepMind, Alphabet’s rightfully revered AI wing, in a statement that AI “It’s all about scaling now”.

Marcus concludes by saying:

“Let’s all foster a field open-minded enough to work in multiple directions without prematurely rejecting ideas that happen to be immature. It may simply be that the best route to artificial (general) intelligence isn’t via alt intelligence.

As I’ve written, I think it’s good to think of Gato as “alternative intelligence” – an interesting exploration of alternative ways of building intelligence – but we need to put it in context: it doesn’t work like the brain, it doesn’t work, it doesn’t learn like a child, it does not understand language, it does not correspond to human values ​​and cannot be entrusted with mission-critical tasks.

It may be better than anything we currently have, but the fact that despite all the immense investment that has gone into it, it still doesn’t really work should give us pause.”

AI alien invasion not coming anytime soon (phew!)

It’s nice to know that most pundits don’t think AGI’s alien invasion is coming anytime soon.

But the fierce debate surrounding AI and its ability to develop human-level intelligence will surely continue — on and off social media.

Let me know what you think!

— Sharon Goldman, editor-in-chief and author

Twitter: @sharongoldman