Thread by @jacyanthis

I'm glad to see leading voices in AI like @sama, @ylecun, and @ilyasut ponder the question of artificial consciousness. We research this @SentienceInst because it's one of the most important questions for the long-term future. I'd like to introduce the topic in a brief thread 1/n

First, the meaning of consciousness is deeply contested among philosophers and scientists. There are 3+ more precise terms:

- thought, typically a linguistic stream of words
- perception, either through the 5 senses or imagination
- sentience, positive and negative emotions 2/n

Even these are contested: Are they behaviors? Processes? Functions? Dualist/nonphysical)? Ineffable? Etc.

I take the eliminativist view that because our terms are so vague, consciousness doesn't exist objectively and we need specific measures instead: sentienceinstitute.org/blog/what-is-s… 3/n

What specific features would indicate sentience or consciousness in AI? Our researcher Ali Ladak wrote a blog post on the specific features that have been proposed. If AI should be called conscious, it will probably be one or more of these features: sentienceinstitute.org/blog/assessing… 4/n

Modern transformers like GPT-3 have few if any of these features. Again, I don't think there is an objective answer to this. Definitions of consciousness like "what it's like to be" something or an "inner listener" are much too vague to answer this. 5/n

https://twitter.com/jacyanthis/status/1284867672333848583

Modern NNs might have "a modicum of consciousness, but probably less than an insect." That's about as precise as we can be with the disagreement, vagueness, and uncertainty.

But the more important question is how we should act regarding the future possibility of sentient AI! 6/n

There are good reasons to think the long-term future will be populated mostly by digital minds, not biological. @xuenay outlined these in 2012. AI can think faster, copy itself more easily, travel at light speed, self-modify, better cooperate, etc. philpapers.org/archive/SOTAOA… 7/n

Holden Karnofsky (@Open_Phil/@GiveWell) has written on this as "digital persons": cold-takes.com/how-digital-pe…

Carl Shulman and Nick Bostrom (@FHIOxford) as "digital minds": nickbostrom.com/papers/digital……

And @RobinHanson (@GeorgeMasonU) as "the age of em": ageofem.com 8/n

There is also a quickly growing field of AI rights: What moral consideration should we extend to AIs? @David_Gunkel, @grok_, @JohnDanaher, @Floridi, @KurtJGray, @JoshGellers, @RomanYam, @GeoChurch, etc. See our recent lit review of this exciting field: link.springer.com/article/10.100… 9/n

@David_Gunkel @grok_ @JohnDanaher @Floridi @kurtjgray @JoshGellers @romanyam @geochurch These are among the most important #Longtermist questions. Follow the #EffectiveAltruism think tank @SentienceInst for new research, and consider applying to our current job openings:

Generalist: sentienceinstitute.org/generalist-ope…

Researcher: sentienceinstitute.org/researcher-ope…

10/n (n=10, for now!)