The Increasing Pace of AI - What Concerns Me

[Warragul]

A recent interaction with Qantas makes me wonder how far advanced Qantas (and probably other airlines are in the use of AI and 'Big Data' with regards to selling or upselling their offerings to travellers.

I'm a WP QFF travelling on a Saver Y fare on a 787 flight but Premium Y is not being sold for reasons unknown. Anyway, as a WP I have selected a seat in the PY section and currently look to be the only one who has pre-selected in that. According to Expertflyer less than half of the Y seats have been pre-selected. So, a seemingly farly light load which should make spreading out across multiple seats as desired a realistic scenario.

Qantas sends me a confirmed upgrade offer which when added to the price I paid for (and near enough to what it is now) that flight leg - as part of return trip booking - is near enough to the business class airfare for that leg (as part of a return trip). So, unless I've had a change of mind about not sitting in Business why would I take this offer (nil extra status credits and points too)?

I know nothing about yield management and the like, so could well be missing half the story, but I think that if all the data - WP who can access the F lounge before the flight for a decent feed etc, half full cabin, already sitting in the PY cabin - was in a data pool/lake/ocean then if they wanted to make an offer that I'd seriously consider then a data model would suggest a more competitive offer.

I'm not moaning because the offer is cheaper, I'm not going to pay or use points for one unless it is ridicously cheap which won't happen, just wondering about possible reasons for the offer amount.

Then again, maybe that' s the idea make what to me is an uncompetitve offer, while hoping others who are estimated to be likely to accept the offer do.

 

[MEL_Traveller]

A recent interaction with Qantas makes me wonder how far advanced Qantas (and probably other airlines are in the use of AI and 'Big Data' with regards to selling or upselling their offerings to travellers.

I'm a WP QFF travelling on a Saver Y fare on a 787 flight but Premium Y is not being sold for reasons unknown. Anyway, as a WP I have selected a seat in the PY section and currently look to be the only one who has pre-selected in that. According to Expertflyer less than half of the Y seats have been pre-selected. So, a seemingly farly light load which should make spreading out across multiple seats as desired a realistic scenario.

Qantas sends me a confirmed upgrade offer which when added to the price I paid for (and near enough to what it is now) that flight leg - as part of return trip booking - is near enough to the business class airfare for that leg (as part of a return trip). So, unless I've had a change of mind about not sitting in Business why would I take this offer (nil extra status credits and points too)?

I know nothing about yield management and the like, so could well be missing half the story, but I think that if all the data - WP who can access the F lounge before the flight for a decent feed etc, half full cabin, already sitting in the PY cabin - was in a data pool/lake/ocean then if they wanted to make an offer that I'd seriously consider then a data model would suggest a more competitive offer.

I'm not moaning because the offer is cheaper, I'm not going to pay or use points for one unless it is ridicously cheap which won't happen, just wondering about possible reasons for the offer amount.

Then again, maybe that' s the idea make what to me is an uncompetitve offer, while hoping others who are estimated to be likely to accept the offer do.

It might be that your company only has an economy or premium economy travel policy. The upgrade to you, as an individual, might be worth taking and laying out of your own pocket.

The system prolly doesn’t know the nuances of every passenger, how they paid for their tickets, and any other factors.

BA does much the same thing with their upgrades, which sometimes represent the exact difference in fare had you purchased business (club) from the outset. There are reasons why it might be appealing… for example the cheap BA fares - both in economy and club europe - are non refundable. If I’m booking a few months out and not sure if i might need to cancel i might just book the cheapest seat to minimise my losses. Closer to departure I might want to upgrade.

In your case, there might be plenty folk who are willing to pay the upgrade, why discount it for platinums? Closer to departure probably plenty bronze members who are on the upgrade list anyway, so nothing lost to QF by trying to maximise revenue now.

 

[Warragul]

The system prolly doesn’t know the nuances of every passenger, how they paid for their tickets, and any other factors.

Thats part of my questioning how far advanced they are with AI etal. They certainly know how the tickets are paid for, maybe not in the Reservation system but in other systems. Thats the raison d'tre of big data and all that, join up disparate datasets or systems to find things hitherto unknown ('valued insights' I think is one of the sexy terms).
The government department I worked at started doing this where it could (some datasets could not be joined for legal reasons) almost 10 years ago.
I've probably over estimated how sophisticated some airlines are in this area.

 

[MEL_Traveller]

Thats part of my questioning how far advanced they are with AI etal. They certainly know how the tickets are paid for, maybe not in the Reservation system but in other systems. Thats the raison d'tre of big data and all that, join up disparate datasets or systems to find things hitherto unknown ('valued insights' I think is one of the sexy terms).
The government department I worked at started doing this where it could (some datasets could not be joined for legal reasons) almost 10 years ago.
I've probably over estimated how sophisticated some airlines are in this area.

But will the AI know you, as employee ‘x’ is only permitted economy travel? Or will it know you are in the C-suite and eligible for business? Or is it just seeing an economy fare, and offering you the upgrade?

Maybe the AI thinks a platinum can afford the full price? Based on other platinums accepting a similar offer.

Lots of data sets, as you say, And they will all be geared to maximising revenue.

 

[Berlin]

I've probably over estimated how sophisticated some airlines are in this area.

Replace this with „most companies“ and you’re closer to the truth!

All that talk about AI this and AI that and once you look closer, it’s some corporate spin together with some good (or bad) old algorithms that are at play (minus a few headcounts of course, because “AI”).

 

[Warragul]

But will the AI know you, as employee ‘x’ is only permitted economy travel? Or will it know you are in the C-suite and eligible for business? Or is it just seeing an economy fare, and offering you the upgrade?

Who knows, probably if it's trained and tuned properly.
Those sorts of things can go in the comments of a booking. Years ago when Mr Abbott was PM, after a storm in a teacup about upgrades he decreed that no public servant travelling on government $ shall travel in First Class ( not USA domestic first class but the real thing), even when it's an 'involuntary upgrade'. Fine no problem with that, my international travel for the government after that had a comment in the PNR to the effect of First Class not permitted.

In my example, I'm self funding so the company policy is involuntary upgrades will be accepted.

 

[Freq Flier 2013]

What I found is that you need to use the right ai system for the right question don’t assume that Claude or Chatgpt is the same. Each one has their own strengths and weaknesses. Also paid subscriptions and various models within them will act differently. Nowadays companies are not just using 1 ai company but 6-7 at the same time. So it’s evolving rapidly.
I use 1 ai for procedure and layout, another one for work specific technical knowledge as it’s only trained on a narrow dataset but very thorough. I use another for graphics and on it goes. Really has changed the way I work.

Will it eventually make me unemployed- sure! But hopefully I retire prior to that.

 

[moa999]

you need to use the right ai system for the right question don’t assume that Claude or Chatgpt is the same. Each one has their own strengths and weaknesses. Also paid subscriptions and various models within them will act differently.

And the right version.

And define what data sources it is allowed to use..
And ask the right questions and hope it doesn't make up an answer.

And remember garbage in garbage out.

 

[Freq Flier 2013]

And the right version.

And define what data sources it is allowed to use..
And ask the right questions and hope it doesn't make up an answer.

And remember garbage in garbage out.

It is becoming smarter. You used to tell it to do all steps from start to end for the ai to do a task correctly. Now you just say the task in general words and it will do most of the work. Meaning your instructions don’t have to be very granular. As ai gets smarter, it becomes more accurate and better to use.

 

[drron]

What worries me is that on at. least 2 occasions the large Ai companies have lost control of their own operations.

Secondly that the very large borrowing yet to come can’t be done and we end up with a global recession/ depression. The Governor of the Bank of England has warned of this.

 

[kookaburra75]

It is becoming smarter. You used to tell it to do all steps from start to end for the ai to do a task correctly. Now you just say the task in general words and it will do most of the work. Meaning your instructions don’t have to be very granular. As ai gets smarter, it becomes more accurate and better to use.

I'd question whether it is getting 'smarter' in terms of producing credible information, as compared to more content. The AI tools are designed to mimic our actions and try and predict what we will ask next, which can come across as being cleverer.

The main failing of the current crop of AI tools, excluding the scientific models being used for health and other research, is they are trained on what is out there, which now includes an increasing amount of AI slop - the AI models are fooling themselves with their own hallucinations. This was a doozy recently EY retracts cyber report littered with AI errors As quoted in the article '...Researchers raised concerns the inclusion of incorrect or made-up information in a report by an established company risks “poisoning the well” by potentially misleading researchers in the future.'

I make use of a different AI tools to help speed up retrieval of information, and create outlines for my work. It's also great at blocking out itineraires and ideas for our travel. But I work on the 80/20 principle, where I expect the responses to have 20% slop in them, and have to cross check to work out where it is and discard it. Sometimes for a bit of fun, I'll get different AI models to critique another model's work. It's interesting what does and doesn't get picked up. Be careful out there

 

[kookaburra75]

Secondly that the very large borrowing yet to come can’t be done and we end up with a global recession/ depression. The Governor of the Bank of England has warned of this.

That too worries me. There's almost ponzi scheme levels of linking what demands might be, where the companies that develop the models, the chip manufacturers, the data centre builders and operators, keep stacking things on top of each other and say it's worth $X. Even the different studies and projections of future workforce demand/changes are laughable, and don't pass the pub-test for investment. As most of it is based on future promises rather than trying to overvalue existing stressed assets aka 2007, I'm hoping the damage will be limited to those trying to hype things up - a Mr Altman or Mr Musk perhaps?

 

[Rayesfeg]

My question for AFF Members is: what will happen when the hits the (such as these days in the ME) , when your only option to contact your airline or TA is via a Chatbot?

No need to worry about that here for loyal QF customers.

Qantas are going to offshore 1000 back-office and corporate support roles so that they can free up some money to "enable the hiring of more customer-facing staff." ... yeah right.

 

[SCM]

Big companies like Amazon have been automating large parts of their operations already even before AI. Legacy/Non-Tech companies have been outsourcing jobs for decades.

None of this is new, but AI actually turns things on it's head a bit. There is now less of an argument to outsource, because co-location and language skills are more important than ever for productivity and moving quickly.

So instead of having say 20 engineers here and 80 in India, it's better to have 30 engineers here, doing equivilent work of those 100 engineers pre-AI, and because they are all together in a single office, and because there are overall less people to collaborate with, less meetings, no timezone conflicts - things can actually be more productive and quicker.

I work in tech, and this is actually the biggest change I see happening. Companies are not interested in hiring remotely now, it's all about being colocated in one place and iterating quickly with AI.

Because the productivity of a single engineer has gone through the roof, it's much better to hire a few good local engineers than a lot of mediocre ones overseas. Saving a bit of money on hiring isn't going to matter when widget company XYZ down the road out-competes you on speed - which is the most critical thing right now.


As for other industries, I suspect AI will have a positive impact on productivity and might shake around the mix of jobs, but overall it won't be as profound as in tech. It is very difficult to adopt new technology - let alone keep up to date with it, the typical worker in the economy is simply too lazy for that. At the same time, many business processes are simply too human-oriented to be easily automated with AI.


There are many jobs that could be automated in the service industry, but it's just not practical because minimum wage labour is still cheaper than complex robots + AI.

 

[drron]

But in Australia there has been no increase in productivity in the last few years. Seems to me that the benefits of Ai might be overhyped.
Our lower productivity here is one of the reasons our real economy is struggling

 

[kookaburra75]

Big companies like Amazon have been automating large parts of their operations already even before AI. Legacy/Non-Tech companies have been outsourcing jobs for decades.

None of this is new, but AI actually turns things on it's head a bit. There is now less of an argument to outsource, because co-location and language skills are more important than ever for productivity and moving quickly.

So instead of having say 20 engineers here and 80 in India, it's better to have 30 engineers here, doing equivilent work of those 100 engineers pre-AI, and because they are all together in a single office, and because there are overall less people to collaborate with, less meetings, no timezone conflicts - things can actually be more productive and quicker.

I work in tech, and this is actually the biggest change I see happening. Companies are not interested in hiring remotely now, it's all about being colocated in one place and iterating quickly with AI.

Because the productivity of a single engineer has gone through the roof, it's much better to hire a few good local engineers than a lot of mediocre ones overseas. Saving a bit of money on hiring isn't going to matter when widget company XYZ down the road out-competes you on speed - which is the most critical thing right now.


As for other industries, I suspect AI will have a positive impact on productivity and might shake around the mix of jobs, but overall it won't be as profound as in tech. It is very difficult to adopt new technology - let alone keep up to date with it, the typical worker in the economy is simply too lazy for that. At the same time, many business processes are simply too human-oriented to be easily automated with AI.


There are many jobs that could be automated in the service industry, but it's just not practical because minimum wage labour is still cheaper than complex robots + AI.

It's an interesting field, and the adoption of AI tools by different industry segments will have different impacts. I've just retired from the tech sector, where the majority of my work was in the strategic consulting space, so I've always applied a very critical eye over what is touted as the next big thing. The key criteria being, will it provide a lasting benefit? After working with and applying Multi-Variable Analysis/Machine Learning/AI systems and tools for the past 20 years, some of it stacks up, some of it doesn't.

Talking with colleagues, looking and evaluating the actual impacts, I don't think we're seeing any improvements yet. In the software field, some companies I've worked with are dialing back the use of AI for coding as they found the additional effort to check what had been put together, and then additional testing due to lack of rigour in the thought process behind the code, is outweighing any benefit. But I do agree that you can change the model as you say, to smaller teams where people are multi-skilled and co-located to create better products, where people engage early and often with each other to make sure the necessary intellectual rigour is developed and applied. Certain AI tools can support that way of working.

Working in and around Canberra, a lot of my work has been with larger organisations, mainly in the public sector. Is this ripe for improvements? No. The productivity paradox is an ongoing issue and AI isn't going to fix it; only make it worse in my mind. Doing some research into white collar worker patterns as part of a white paper earlier this year, in the 1970s the split was 60% work, and 40% work-about-work i.e. anything that didn't support output e.g. timesheets, internal meetings etc. By the start of the 2020s the split has moved to 40% work and 60% work-about-work i.e. even more internal meetings, pep talks, mindless emails, trawling internal portals trying to find 'something'. The AI movement says AI tools will help write emails faster (but did you need to read/send that email in the first place?), send AI bots to meetings in your place to summarise what happened and generate a list of actions (what if no-one actually attended and sent their bots along in their place?), collate information (which then has to be verified due to it containing AI slop). Will any of that move the dial to improve the current 40/60 split, or just replace part of the work-about-work segment? Or even worse, move the dial the wrong way? And then there's the underpinning cost element. Are AI companies going to reduce the cost of their products or services over time once they are considered essential? Look at how Microsoft monetised Windows and Office once it had a position of 'power'.

SInce the 1970s we have been told in implementation of 'technology' will save time, improve our lives - we were asked what would we do with all of our spare time. But the reality is that hasn't happened. Many tools and systems I have seen are often, a "solution in search of a problem." I'm hearing much the same hype around the AI LLMs eg ChatGPT, Clude, Gemini, but no-one has put together how they will evaluate the supposed improvements, it's all done by magic. At the most recent place I worked we introduced AI Usage Briefs, where the why/what/how use of AI Tools had to be documented and signed off - even if the answer was 'none'. Partly to provide some comfort to the client what would or wouldn't be used, and also to limit our liability to a consultant going rogue and doing something dumb with client data and/or delivering a dodgy report.

Where we will be in 5-10 years time will be interesting. My biggest worry is that entry level jobs for young people will dry up, while companies make do with AI Tools. That will 'hollow out' those companies in terms of mix of age, experience, and ability to bring in new skills, which is something AI tools can't do. They're stuck with what they are trained on. That will be a major structural problem.

[AI Usage Statement: no AI Tools were used in the production of this post]

 

[AuSammy]

So far, most AI solutions have been absolute slop and never solves the customer service issue you are having.

My experiences

ANZ none of at least 6 requests for assistance or a task to be performed were understood correctly, answered correctly, did not provide a misleading answer - total and utter garbage.

Westpac - only around 3 queries - exactly the same response as ANZ AI

ChatGTP

More than 12 months ago now but at the physio the receptionist said some health funds now have digital cards you can put in your phone wallet so you don’t have to carry the card around.

On arriving home I searched the QF health insurance site and couldn’t find any reference to a digital card - so I asked ChatGPT if it existed and where to find it. I was presented with

“Yes its definitely available followed by a detailed lengthy procedure on how to log on go through three menus and select digital cards and download it.

I went through the app and none of the menus quoted and options existed.

Strange, maybe I haven’t got the latest version of the app.

Rang QF health - no we haven’t implemented digital cards, yes you have the latest version if the app and you are correct those menus do not exist, where did you get that false information from.

ChtGtp had completely invented a fabricated story without a single element of truth. [in the last 12 months QF health have released digital cards but they certainly didn’t exist at the time of my query]

 

[MEL_Traveller]

I’m looking forward to the benefits of AI. Travel related… air traffic control: what is the most efficient use of runways, aircraft type, push back clearance, zippering of aircraft on hold and coming in to land, etc.

In the medical field,,. to do a second opinion after the doctor or consultant has reviewed the case. Anything missed? And for cases which sometimes remain undiagnosed for years… scooping up all the anecdotes from around the world, reddit, facebook - you name it - to try and find a diagnosis.

They tries it in the law field… bit of a failure! But that was because the lawyer handn’t checked the work.

As an aid or assistant, i think it’s going to be invaluable, I wouldn’t trust it just yet as a sole decision maker, unless there were simple and easily accessible avenues for appeal.

 

[MEL_Traveller]

ChtGtp have completely invented a fabricated story without a single element of truth. [in the last 12 months QF health have released digital cards but they certainly didn’t exist at the time of my query]

According to a search I did last year, China Southern is *absolutely* a member of skyteam, star alliance and oneworld! Concurrently even!

I see some references have thankfully been corrected!

 

[AuSammy]

According to a search K did last year, China Southern is *absolutely* a member of skyteam, star alliance and oneworld! Concurrently even!

I see some references have thankfully been corrected!

Similarly list all the BEVs available for sale in Australia under $65,000.

“There are only these three” - Absolute