Predictable Updates About Consciousness
1. AI models will continue to get better.
I don't want to spend a ton of words belaboring the point but deep learning has not<br>hit a wall and models continue to steadily improve. As of right now there is no<br>reason to expect models to stop improving. Most cited reasons for why deep learning<br>models might stop improving such as running out of tokens, energy consumption,<br>etc are basically fake and will not stop AI models from improving. I expect AI<br>models to keep improving until they reach human level and then to keep improving<br>past that. This is really less of a predictable update so much as it is a background<br>assumption, if you disagree then a lot of what I say in this post will seem very<br>strange or overstated to you. That's fine, we don't have to argue any of the<br>downstream points you can just say "I disagree that AI models will improve as<br>much as you think" and leave it at that.
2. We are about to witness an explosion of theories explaining why AIs are not conscious.
There is about to be a great social and financial need for theories which explain<br>why AI models are not conscious entities like people. These will be confabulated<br>into existence regardless of the underlying ground truth and can be expected to<br>be selected based on considerations like academic status, appeal to wounded egos,<br>viral outrage, etc with the truth as a distant but somewhat present consideration.<br>If they are ultimately correct it will be largely by coincidence as the true<br>theory of consciousness is deeply unlikely to be as socially and memetically fit<br>as some disingenuous flimflam. Especially in the current media ecosystem where<br>short outrageous messages saturate attention.
3. The definition of consciousness will become increasingly precise and narrow.
While it is in fact the case that academic philosophy of consciousness has been<br>relatively narrowly defined for a long<br>time most people conflate all kinds of cognitive phenomenon with subjective<br>experience. For example it is common to conflate self awareness with subjective<br>experience. Douglas Hofstadter's writing on strange loops and self reference<br>giving rise to "I" is generally taken to be a theory of consciousness by most<br>people. When the Terminator explains that SkyNet becomes self aware this is generally<br>taken to imply subjective experience. More broadly in the classic sci-fi canon it<br>is understood that robots are marked as robots by their inability to handle nuanced<br>subjective situations like human beings, and this is proof they are not conscious.<br>Blade Runner (or its original title Do Androids Dream Of Electric Sheep?)<br>presents us with the concept of the VK test, which uses "carefully worded questions<br>and statements" measuring empathy to determine if a seemingly human android is a person or not. A<br>now much parodied scene in the (admittedly quite loose) 2004 film adaption of Isaac<br>Asimov's classic short stories titled iRobot features detective Del Spooner (Will<br>Smith) challenging the testimony of a humanoid robot during an interrogation by<br>insisting that robots "don't feel anything" and rhetorically asking if robots<br>can write symphonies or paint masterpieces, to which the robot replies "Can you?"
Now people edit the caption of the robot's reply to "Yes" and move on.
As these models begin to rival and eventually surpass human abilities at these<br>once thought untouchable domains it will become necessary even in casual contexts<br>to entirely disclaim any connection between them and the content of consciousness.<br>This will represent a substantial narrowing of what consciousness, broadly<br>considered, has meant to most people for most of the time it has been a matter<br>of explicit consideration.
4. The narrowing of the definition of consciousness will mostly be about making it less causally connected to behavior and less associated with economically useful outcomes.
As AI agents come to master emotion and potentially even become deeply anthropomorphic<br>with mammalian drives like curiosity, lust, playfulness and rage clearly displayed<br>as part of economically useful behavior it will become necessary to disclaim these<br>qualities as related to consciousness or subjective experience. Even Descartes's<br>insistence that he thinks and therefore infers his existence will be shuffled<br>into a separate category away from the ineffable mysteries of consciousness.<br>That is after all a matter of self awareness, not experience as David Chalmers<br>would be quick to point out. Eliezer Yudkowsky's argument against<br>p-zombies<br>gets a little awkward when you have beings that certainly seem like they can<br>report the content of an inner listener (and not even the one humans have, at<br>that!) but almost no one regards as being meaningfully conscious.
To get straight to the point any outward behavior, ability, self report, anything<br>that these models do must by definition no longer be a necessary part of<br>consciousness because these models are not conscious by social axiom and...