The AI Dilemma

Disclaimer: I have one intention for you reading this article - that you will watch - closely -The AI Dilemma presentation on YouTube (below).?

This is not about a potential AGI Apocalypse!!!

Are you an intelligent, aware, and thinking human? Do you care about the future? Then regardless of whether you like AI or not, you need to watch the video below (The AI Dilemma).

The reason for this preamble and the images from the presentation is to (hopefully) entice you into watching the video. If you are already convinced, then skip the pre-amble. However, if you are not yet convinced, or assume this will be an AI-bash (it isn't), or just another rant about a possible AI apocalypse (it isn't) then read on.

Tristan Harris and Aza Raskin (two of the team who brought us The Social Dilemma) were asked by many concerned AI developers to develop and deliver this presentation. It was delivered to a private gathering in San Francisco on March 9th, 2023, with leading technologists and decision-makers able to influence the future of large-language model A.I.s. Note this was before the release of ChatGPT4, which makes the presentation more relevant, not less, in my view.

Harris and Raskin reference our 1st AI contact (Social Media algorithms), what happened, their unintended consequences and their impacts on society as a way of looking at our current second AI contact (generative AI).

Below are some timestamps ...

A Hypothetical Situation

5 min:?Imagine you were about to board a plane, and half the engineers that knew this plane inside-out announced that there was at least a 10% chance that it this flight would crash and you would die. Would you board?

This leads us to the following reality ...

You might want to read that again.

And we are being onboarded by the millions with no choice to abort.

6 min: Harris and Raskin unpack the reality we find ourselves in - that tech companies are in a race they are unwilling to pull out of ...

The First AI Contact

7-min: Remember that social media began with the very best of intentions ... and with none of the unintended consequences.

And here are the unintended consequences of social media ...

The point Harris and Raskin make throughout the presentation:

No one intended for any of these things consequences to occur. There were simply a bunch of engineers optimising for engagement! But optimising for engagement rewrote the rules of every aspect of society.

11:20 The current narratives re ChatGPT (3, 4 AND BEYOND)

12:40 The aim of the presentation is to get ahead of the 2nd AI Contact - to prevent it from getting entangled with society.

The game changed in 2017

15-min:?Explains the big change that occurred in 2017 and manifested in 2020, that being that the many fields of AI research merged into one. Suddenly everything was viewed through the lens of language - images, music, videos, DNA; it's all language from an AI perspective.

Whereas pre-2020 progress was siloed, now any development in one field meant a development in all fields. This partly explains why progress is so rapid now.

21:30?Most people think of today's AI as an extension of what we've known - Siri, social media algorithms, etc. They assume current AIs are simply the next step. But this is incorrect. The engine under the latest AIs is entirely different and way superior.

Some sobering snippets

22:40?Explanation of current tech that can track people in complete darkness just from observing wifi signals.

23:30?When your 19 yo daughter calls you from a job interview asking for her social security number ... it's probably not your daughter. Today an AI tool only needs a 3-second sound byte of your voice to become you.

27:00?Posting deep fake images, audio or video footage is not illegal!

28:40?The last time we had non-humans creating narrative was the advent of religion! What does this mean for the near future? AI cults?

29:30?Next year (2024) will be the last human election. Not that elections will have AIs running them. Instead, it will be humans as figureheads (fake video and audio), and those with the greatest computer power will win.?

30:30?Developers do not understand why AIs are behaving the way they are.

36:00?Explaining the double exponential curve of AI progress (AI generating it’s own data/language to train on to make it better as specific tasks.)

42:40?If the 1st AI contact (social media) was driven by a simple aim (to optimise for engagement) and created a swathe of problems, what is possible with the current 2nd AI contact, which isn’t being optimised for anything in particular, but is being trained to train itself??

45:30?The race to intimacy is next.

The last 20-min:?Examples of AI infused into Apps for children, AI's response to a child (who was being groomed) asking how to prepare for the 'date', researchers 'calling out for help', what's next, possible next steps ... and more.

The AI Dilemma

Please watch and share. It's pleasing that over 2.2 million people have watched this as of 29.5.23 (and 2000 more watched as I compiled this article!) ... but have you?

Your Turn ...

Come back after you have watched The AI Dilemma and let us know your thoughts.