My exploration of A.I. using Chat G.P.T.

The case for

I used Chat G.P.T. to draw the above image. The instructions were very straight forward, and the image output astonishing – with a few exceptions, it just about completely matched the image I had in mind of a hydroelectric turbine at the end of a power station penstock. The water trickling along the floor was a bit much given that the intake gate was close, and the turbine itself did not have large enough, or sufficient blades to adequately process the volume of water that would been passing through.

The time it took me to write the instructions and mistake proof them: 3 minutes. The time it took Chat G.P.T. to write its interpretation: 1 minute. To turn that into a significantly detailed schematic when prompted: about 2 minutes.

I also used Chat G.P.T. to draw me a schematic of a waste to energy plant. Again, a very quick turn around, though more detail was required. Perhaps all the more impressive because it annotated the schematic, in addition to drawing it.

These were part fun/part serious attempts to see how the programme works.

Others have tried using Chat G.P.T. and similar to create things like holiday itineraries. It is not something I have considered, as I want to be in full control my itinerary – and be fully responsible – for any downfalls created, instead of a creation that would probably find a way out of blame for any problems. Still more have apparently designed houses, had constructive arguments with it around legal frameworks. I have engaged it in talking about American foreign policy, which was surprisingly productive given that A.I. is not human.

The case against

All technology has its uses. All technology has its abuses, and it is in this section I explore some of the abuses. There will be – as former U.S. Secretary of Defence Donald Rumsfeld said – “the known knowns, the known unknowns and the unknown unknowns”. He was talking about threats in Iraq during the sectarian violence, and was describing the ability of the U.S. military to identify and deal with those threats.

There are A.I. threats that are well known to researchers and analysts. There are threats that we know exist, but know nothing about. And there are the ones that as yet have not caught the attention of researchers.

The ability to create disinformation through various media now being made possible by A.I. is scarily large. And I do not just mean in terms of text messages or articles promoting information known to be wrong. On Facebook on an almost daily basis I see images of volcanic eruptions that have clearly been doctored; a supposed chasm in the ground, somewhere in China kilometres long, and ending far too abruptly and being far too orderly for a natural ground rupture, is apparently meant to be a new rift valley opening. Yet Facebook, despite being told innumerable times that it is disinformation, never finds the creator guilty.

The ability to create degrading and harmful images and video of people is another threat. Whilst the A.I. systems attached to the main browsers such as Microsoft Edge and Firefox are quite strict with inbuilt rules about what can constitute harmful material, there are several other A.I. systems with few or no rules. They include, but are not limited to Venice, and OpenAI. With Chat G.P.T. it is relatively easy to trigger one or more of its rules about content generation. A search on Google about rules around content generation for the lesser restricted A.I. programmes suggests that it would be very easy to take a photo of someone, and with a few clearly worded instructions, completely strip the clothes off them.

The author’s take

Initially I was not really sure of what to make of A.I. Automation is coming for many jobs in New Zealand and around the world. Most people would have no problems serving themselves at a supermarket check out. Would they be okay though, as some technology enthusiasts/libertarians(?) have suggested that pharmaceuticals be dispensed by robots? Would they be okay with self driving vehicles having a major role on our road given the frequency with which there seem to be accidents?

As with other parts of the A.I. spectrum, I see no point in trying to point blank block children from using A.I. – they will simply look for ways around any such attempts. The closest I can imagine, comes through registering children on cellular phone networks, which typically ask for ones age when doing so. If a person is registered, they will be asked to present evidence of self, such as a driver licence or passport.

I asked Chat G.P.T. how it thought rules should be enforced across children and teens. Here is its take verbatim. Of the technological challenges posed by other forms and uses of A.I., that is a conversation for another day.