[LINK] ChatGPT

[David]

G'idday Roger,

There's so much potential discussion in your Paper "A Re-Conception of the Field [of AI and Robotics, 2023]" http://www.rogerclarke.com/EC/AITS.html#RAI that any comment which covered the ground properly would be too big for a post to Link (and it would take me too long to get together!).  You could easily run a one-day professional seminar on the subject along the lines of those run by Sydney Ideas at SydUniv.

Of one thing I'm confident: public, legal, and legislative awareness of AI urgently needs to be improved, and "a re-conception of the Field [of AI and Robotics]" seems right on the mark.

Neural networks have been around since the late 1960's so "AI" isn't anything new.  However I suspect most people's conception of "AI" is still so vague as to be quite misleading, and when this extends to those in the legislature and judiciary we have a problem.  The very expression "artificial intelligence" is unfortunate because it suggests a computer which works rather like the human brain, capable of human judgements and emotions.  This probably isn't helped by a statement attributed to Elon Musk I noticed somewhere recently to the effect that he couldn't be sure the computers in Tesla cars were not sentient to some degree.

So... presentation & emphasis depends on the audience.  But I suggest putting more initial emphasis on the differences between any AI system (at least for the foreseeable future) and the human brain.  The big one of course is consciousness and how it arises, a problem for which there's not yet any plausible theory as far as I'm aware.  But there's no getting away from the fact we're conscious beings, with a complex range of social and intellectual responses and a genetic memory of our relations with other humans.

But an AI system is a fast, multi-factor correlation engine: if the sensors show this pattern, do that.  (I think a Linker pointed out some time ago that the CAPTCHA challenge commonly employed to demonstrate a person is not a robot is used to train the AI computer in self-driving vehicles.)

In summary, I suggest a general "re-conception" of AI has to begin by clarifying why it is in no way like the human brain.  "Re-conception is a strong word so it needs a strong response IMO.  I think we should drop the term "AI" altogether if possible; "autonomous system" is better but probably not much clearer to the non-specialist.  Maybe some Linker could suggest a colourful phrase which would catch on in the public mind...!

_David Lochrin._


On 31/1/23 19:04, Roger Clarke wrote:
> Thoughts on the criteria to use in deciding which level of device autonomy is applicable to any particular situation are in:
>
> s.2.2 Drone Control (a) Autonomous Control [Drones specifically, 2014]
> http://www.rogerclarke.com/SOS/Drones-E.html#DCA
>
> s.4.1 Artefact Autonomy [AI generally, 2019]
> http://www.rogerclarke.com/EC/AII.html#TAA
> in particular the para. just after the Table
>
> s.5.1 Complementary Artefact Intelligence
> http://www.rogerclarke.com/EC/AII.html#CI
>
> 11. A Re-Conception of the Field [of AI and Robotics, 2023]
> http://www.rogerclarke.com/EC/AITS.html#RAI
> proposing the primary focus be on decision support, incl.:
> -   Complementary Artefact Intelligence
> -   Augmented Intelligence
> -   Complementary Artefact Capability
> -   Augmented Capability
>
> Critical consideration of the above much appreciated!
>