Beoz

Correct. That's why people automate or don't as the case may be. Invest X to receive Y. Common sense. That's why we very rarely automate to cut human costs. The returns are not there.

Staff, just ticking over. All good in a good market. No so good in bad markets which is when you cut. Nothing to do with automation.

Beoz

The above is not just for sales, its for every day business.

With the above you could send your clone bot to the meeting while you sit on the beach. Sooner or later our clonebots will be doing it all for us.

However, there are a couple of vital things missing here.

Creativity - what happens when that business deal is not going your way, and you need to get a little creative, think a little outside the box. Will our clonebots have that ability?

Learning - sure, you can teach them stuff, some stuff, but not everything. Experience for starters. Been there done that for 25 years. Learning by making a mistake. That type of stuff.

Yeah I don't think so.

Shard

Probably about 20 years on fax. Agree re. telesales, was just responding to Gary on capabilities.

Shard

Nope, nobody is going to be sending 1950s style tin robots to board meetings. The whole model of doing business is going to change once machines are interfacing with each other as defaults. Small groups of humans will still make the tough and creative decisions, perhaps in a boardroom, perhaps on the beach, but there will be no need for armies of office staff to collate and present data. Think real-time what if "oracle". Also, AI is no longer "taught stuff" it's guided to huge datasets and does the learning itself. It's kind of black box, which is both a strength and risk.

GarryP

Do you have a Google Home? It's not totally there, but then again, neither are most salesmen. A judicious amount of ignoring people and ploughing on regardless will deal with many issues.

Or while you sit in the gutter, jobless.

Oh, you mean lying? Yeah, AI will be able to lie as well as a salesman.

That's EXACTLY what they do learn, from the pattern matching of countless hundreds of thousands of situations and outcomes. That's the whole point, recognising and pattern matching what has been done before. Real creativity is the issue, but as we know, that's rarer than rocking horse sh*t.

Beoz

I think we have very different ideas on what selling is and what century we are applying it to.

Patterns don't apply to those who can sell.

Wol

Moving on.....
I see today that the UK defence department is to buy fighters that are autonomous since they will not put pilots at risk. But they will be able to have pilots if a task needs them.
Leaving aside the questionable assumption that an effective autonomous fighter aircraft is going to be possible within the entry timescale (2034 IIRC), what genius thinks that having the option of putting a pilot in the thing is viable? Much of the weight of having a pilot is involved in ejector seats, life support, emergency equipment etc etc.
The so-called AI industry is leading politicians by the nose over all aspects of autonomous systems.

GarryP

OK, this is something that stretches back at least 20 years. To understand it, you have to understand the high ups in the RAF. They are generally promoted from the ranks of pilots etc. and they view pilots as irreplaceable, and autonomous aircraft as an anathema. The 'unmanned but can be manned' idea was developed to sell them on an idea of the 'next generation of combat aircraft' where those pilots still had jobs (even though nobody takes it seriously). IIRC JSF and Eurofighter were originally designed to have autonomous modes like this, which these ex-pilot types killed off. You see reference to 'swarming', where the idea was these pilot were going to tactically direct the autonomous instances (fat chance).

Reality is there is no place for the Top Gun pilot types - the drones fielded by the army put paid to that - but whilst they have control of the purse strings, BAE will continue to try and sell them on manned fighters. This "Tempest" thing looks like something that BAE have be trying to get off the ground for at least 20 years - in fact it doesn't look like it's moved on in that time - seems similar to what they were suggesting for a Tornado replacement.

In fact, a little research finds :

https://ukdefencejournal.org.uk/uk-u...alled-tempest/

The problem is not creating an autonomous aircraft, it's getting them to accept that they won't be manned.

Shard

There's also a huge moral quandary regarding the development of autonomous weapons. Governments may want to keep skin in the game (literally!) as an arms race in killer bots is not a desirable outcome for the species.

Beoz

Isn't there already autonomous fighting aircraft. Maybe not with a jet engine but ........

Wol

Thank you for that.

I think that fully - unmanned - autonomous fighters will be a long time coming, just as will be fully autonomous cars on ordinary roads. Same with civil aircraft: pilots are able to contain mechanical and operational issues which the most sophisticated AI has no chance of addressing, and the human and financial penalties of losing a civil aircraft and the operational cost of losing fighters is unsustainable. Perhaps in a century, if we get that far which I doubt.

GarryP

Basically they are already here. The only reason you don't see them is because of doctrine - they would be much more effective in a whole host of dimensions. UCAVs are simpler than autonomous cars to engineer, and even they are almost here already.

I can tell you one country that won't be wondering about ethics.

Image emerges of China’s stealthy Dark Sword UCAV | Jane's 360

Wol

I don't know your own background, but I think a lot of people - and industry and politicians - are confusing simulated AI with actual AI. By "simulated" I mean what we see at present - massive programs that present solutions that their writes can't see how they arrive at it. One can go all metaphysical and argue about sentience, self awareness, intelligence et al, but the fact remains that present so-called AI are nothing like intelligent in a way that we understand.
With autonomous fighters, for example, you can only code up to a certain level of sophistication, to deal with thousands of different scenarios. When something occurs beyond that level - say, for example, the thing returns to its base which has been obliterated - it would have to have orders from the base - but communications would be a very vulnerable problem. Much the same with self driving cars, which are arguably orders of magnitude more complex to program but which can at least stop and wait!
One car "stopped, waiting for assistance" is not a problem - but when (if) millions of self driving cars do the same thing......
I'll believe it when I see autonomous cars on the streets without any manual reversion - and I don't mean the little bugs that are being tested in very limited areas, but on all sorts of roads.

GarryP

To me the whole "but is it really artificial intelligence" thing rapidly goes up it's own arse, particularly when the philosophers stick their oar in and try to say nothing for 3 hours straight because they love the sound of their own voice. I like to keep it simple - if it looks like a duck and quacks like a duck; it's a duck. Particularly if it does something useful. You could rightly call the current deep learning stuff pattern recognition. But when it displays more 'intelligence' than half the human race, I'm happy saying it's intelligent, even if it's just layered neural nets.

Looking at combat aircraft, you can apply the gold/silver/bronze viewpoint to their operation - strategic/tactical/operational. Now no combat pilot plays at the strategic level, control is very much top down and it's outside their pay grade. And tactical tends to be done by mission planning software, if for no other reason than that coordinating all the required elements and avoiding complex defences is not something you can or should be trying to do whilst flying the aircraft. Automation means that you can stuff that mission planning software onboard the aircraft, interfacing directly with the rest and probably resulting in a more agile and responsive, whilst still coordinated, capability.

Which leaves us with the operational.

Now, you'll take onboard that the aircraft is already really flown by the automation, and that generally what the aircraft is capable of is beyond what the pilot can sustain. Also the 360 sensors and fusion to create a recognised picture is automated and you then have to try to hammer it into the pilot's head through the UI.

Upshot is that there is a VERY narrow set of required capabilities that the pilot can play in - basically tricks to try and outwit the opposing human pilot/air defence/etc. But again, these are exactly where the pattern recognition is enough to be able to do as well, if not better, than the human pilot.

Taking your example, return to a base that's not there, it's already process bound - they are taught what to do, and automation can do the same. If you had something really out of left field happening then you do what pilots do anyway, you abort and return to base. That's what the COO says, you don't get creative.

There's a reason autonomous has been seriously planned for over 20 years - it's been doable for that long, and advances just make it more capable (something the human pilot isn't).

scrubbedexpat091

We still use fax machines at work so operations can fax the flight info down to us on the ramp since ramp office and ops office is nowhere near one another. Unless the company issues phones with data (they won't) I can't think of any other way to obtain the information currently. Suppose over the phone, but eh fax is easier then you have an actual copy.