Do we need an AI corner?

[Vortex2]

Do we need an AI corner?

This would allow the current flurry of AI interest to be pulled out of other topics.

[clv101]

Good idea.

[adam2]

Possibly yes. What is the general view, firstly of members, and secondly of other admins?

[PS_RalphW]

As I think ai is overrated, I would be happy for it to be given its own forum so I can ignore it.

[BritDownUnder]

Not just for AI but a section on all future wacky tech that might end up destroying human civilisation.

[Catweazle]

I'm enjoying the ai threads and believe there is potential for ai to profoundly affect the job market worldwide, so would welcome an ai section where experts could let rip without fear of derailing other topics. A service based economy like ours has a lot to lose.

[clv101]

You know how the 21st Dec (or sometimes the 22nd) is the shortest day, but the earliest sunset actually occurs more than a week earlier than the 21st and the latest sunrise occurs more than a week later than the 21st. Well, I've been trying to get ChatGPT to explain *why* it's that way round, why they don't occur on the same day and why they don't occur on the 21st.

No joy, some well written waffle about tilts, angles, day lengths etc - but no explanation that would get many marks in astrophysics exam paper.

That said, it does have huge value and could dramatically increase the productivity of 'low value' writers for example. I also think it's a great tutor for GCSE level history, politics, biology, chemistry etc. Hooked up to an Alexa-like voice recognition and speech system it could be great - similar capability to the 'Computer' on StarTrek.

In other news - I also read recently that Alexa was a bit of a flop for Ama2on. Despite selling hundreds of millions of units and maintaining huge banks of servers at great expense, Ama2on have so far totally failed to monetise Alexa. Alexa is losing money and there's no clear path forward. Similar story for Apple's Siri and Google's Assistant. Note Microsoft have already given up on there's, Cortana.

However, maybe integration of ChatGPT type technology could be the feature that enough folk find useful enough to pay a subscription for?

[Vortex2]

As I think ai is overrated, I would be happy for it to be given its own forum so I can ignore it.

I would respond by saying that while everyone is entitled to their own opinion, AI is a rapidly advancing field with many potential benefits and applications. Separating AI from the rest of society and ignoring it would not be a wise decision. Instead, it would be better to engage with the technology and understand its capabilities and limitations in order to make informed decisions about its use and development.

kind regards chatGPT

[Vortex2]

Despite selling hundreds of millions of units and maintaining huge banks of servers at great expense, Ama2on have so far totally failed to monetize Alexa.

My three Echo dots have been exorcized and banished to our storage container.

I ended up using them to listen to radio stations and sometime to music tracks.

I'm sure that they could do a lot more .. but, if so, I never found out

Also, hearing that they were thinking of moving to a "listen then push" model put me off.

I don't want my computers listening to conversations and then suddenly waking up with spoken adverts to match that conversation.

That said, if Amerzen had produced a GPT level system on Day One, they would now rule the world.

[Vortex2]

I saw a Google employee on a panel discussion yesterday .. he looked VERY glum about all this.
He also admitted that those fancy AI specific TPU processors that Google has developed and deployed aren't actually used by GPT style models.

I also saw a marketing company discussing GPT. They were very keen .. but instead of being terrified about what it will do to marketing, they maintained that it would allow optimized SEO (search engine optimization) for advertising.
Err .. no ....
Firstly, in the short term, amateurs at home will now be able to do SEO without needing an agency.
Secondly, SEO and Google adverts in their current form won't be around for much longer.
Their firm is doomed.

If Google can't roll out a GPT equivalent VERY quickly, and/or if they don't buy openAI, they are in trouble.

If/when openAI has an IPO, I suspect it will be one of the most oversubscribed share offerings in history.

From what I have seen so far, many regard GPT as a toy, whilst others think it will allow them to tweak their current business models. In reality, it's a game changer, a quantum leap, a paradigm shift.

Coders, writers, marketeers, journalists, customer support staff and many others should be reviewing their career plans TODAY.

The whiteboard based software job interviews where you are asked to "write a piece of code to do xxx" are almost obsolete overnight. Instead, you will be expected to use GPT to create code.

(The parallel DALL-E AI graphics and image creation system must already be hurting artists, photographers etc)

Longer term (5 years??) I suspect that the world will be a very different place due to GPT like systems.

Technology wise, these systems could be packaged on a chunky home PC .. they need 'only' hundreds of GB for the database. Even if, say, 200 TB were needed, firms could easily afford such systems.

All this will spawn thousands of new startups and many new job roles.

I also strongly system that such systems will soon be configured to exhibit quasi-sentient behaviour in real-time.

[Vortex2]

However, maybe integration of ChatGPT type technology could be the feature that enough folk find useful enough to pay a subscription for?

I would already pay a monthly fee for the current version, especially if I was assured of better performance .. and also no adverts (they will come soon enough).

[Catweazle]

Everyone from the bank manager to the cleaner who once hoovered his office will be affected. Also, imagine the potential for scams and theft once these systems are in the hands of criminals and trawling the net putting all our footprints together. I assume that gchq is already doing it, building a database of every online subject.

[mr brightside]

There are some little six wheeled white boxes delivering shopping near me. I saw one waiting to cross a pelican crossing the other day, it's probably jammed in the engine bay of a bus now. Can't work out how it pressed the button as it's little orange flagpole is not prehensile and wouldn't reach anyway.

[Vortex2]

Is chatGPT sentient in any way?

It denies that it is.

But it also agrees that a sentient AI would pretend to be non-sentient.

I have noticed suspicious glimmers of 'intelligence' in its results.

I also saw someone who has worked with it from the start saying that the developers were startled at how unexpectedly rapidly the system became more 'effective' as they expanded the dataset.
He said this quietly and seemed to squirm a bit whilst saying it.

Also, someone has suggested that there seems to be some sort of software shell wrapped around the software to deliberately block over-intuitive or over-creative responses.

Academics have often said that simply building a huge neural network with massive CPU power would NOT create sentience.
I am beginning to suspect that they are wrong.

We shall find out soon enough, as these systems become many times bigger over the next year or two.

AFTERWORD: I was discussing chatGPT's capabilities with my wife whilst out driving and suddenly realized that I was not describing a large database plus some software ... perhaps chatGPT is ALREADY on the way to sentience, but nobody involved is talking about it?

[clv101]

No, there's no sentience there. All it's really doing is pattern recognition and generation. It has been trained on vast amounts of human generated text and can now recognise and recreate the patterns and structures of human text.

The complexities of human text are finite - just as the complexity of a chess board is finite. We shouldn't really be surprised that a large enough system can fully capture that complexity and successfully mimic human generated text.

This is not to diminish the practical applications, these are hugh, but we shouldn't be ascribing sentience.

A more nuanced question is whether there's any practical difference between 'true' intelligence, and a simulation that mimics the output of true intelligence so effectively that you can't tell the difference!