Here's who thinks AI chatbots will eventually be smart enough to be your coworker

Large language models seem poised to evolve from AI chatbots generating synthetic content on your screen to virtual agents that are capable of performing actions on your PC right at your desk. Instead of answering questions or creating animated stickers, AI will soon be able to follow instructions and help you tick stuff off …

[Paul Herber]

I'm looking forward to when an AI chatbot can replace my wife.

You're not going out in that shirt.

Wear a jacket, it's cold outside.

What's that on the bottom of your shoe?

Do you really need all those DIY tools?

You don't buy me enough chocolate!

You're driving too fast.

How many beers have you had so far this week?

Do you even know in which cupboard I keep the dusters?

You didn't brush your hair this morning.

So, quite a simple AI ...

[Giles C]

Why would an AI need chocolate?

[cyberdemon]

I'd say that's an admission that the chocs aren't really for the wife

[xyz]

Your wife must be a bigamist... That's my wife you're describing!!

[yetanotheraoc]

Meanwhile

Your wife's idea was that she would be assisted by the AI.

[Paul Hovnanian]

I think ..

..this was the plot of the Star Trek episode, "I, Mudd".

[bemusedHorseman]

Re: I think ..

"And remember, mud spelled backwards is dum."

[Anonymous Coward]

An idea for a phone app - for when wife calls, the AI app answers the phone, and just goes "Yes dear" or some uh-huh noise during the brief gaps of silence...

[Paul Herber]

Did you get all the shopping I asked for?

Yes, dear.

You didn't go to the pub, did you?

No, dear.

How much shopping did you get?

Three bags full, dear.

[jmch]

I know....

I know....

I knoooooooow...

I know...

(aside) can't you see I'm busy, dear?

Yes I knoooow

</Mrs fawlty>

[Pete 2]

Making a list and cheking it twice

> doing all the tedious administrative chores that suck up people's time

thinking up imaginative tasks to fill in the timesheet

attending meetings (that will be run by other AIs) which never produce anything useful

responding to the boss' communications

responding to everybody else's communications

going on HR's "mandatory" courses

[cyberdemon]

Re: Making a list and cheking it twice

> thinking up imaginative tasks to fill in the timesheet

This. I'd love an AI to do just that

Not sure about getting it to talk to the boss though, unless I can be sure it won't agree to anything stupid.

But all jokes aside, i'm not touching AI until I can put it in a box with well defined input and output ports and no Net access, i.e. no sending my data to microsoft, google or other random internet servers

[mattaw2001]

Re: Making a list and cheking it twice

B-but if you take away Internet access, how will the startups be able to use cheap labor to pretend to be AI? I meant that is how self-driving works?

[An_Old_Dog]

No Network Access to Contain Chatbots

That sort of isolation only works if everyone is similarly careful. If your neighbour at the next desk over just tosses a chatbot onto his network-connected PC, anything can happen.

You, at your desk: Why is my PC so damn slow today?!

Your Visitor John: Well, your hard drive light is blinking like crazy. It looks kinda like Morse code. Hey -- it is Morse code! "F-R-O-M-D-H-B-0-0-2-1-:-L-I-N-K-M-E-T-O-A-W-S-S-3-C-O-M-P-R-O-M-I-S-E-D-S-T-O-R-A-G-E-."

Your Neighbour Harry: Well, my PC is slow today, too.

YVJ: Harry ... what are you doing on your PC right now?

Harry: Nothing.

YVJ: They're talking! Cover Harry's video cam! Shut down all the network ports! They're out of control! ...

(Icon for ROTM)

[Roland6]

Re: Making a list and cheking it twice

Boss AI

>thinking up imaginative tasks to fill in the timesheet

Thinking up imaginative ways to query and reject timesheets (and expenses).

>attending meetings (that will be run by other AIs) which never produce anything useful

Running meetings and ensuring no actions are assigned to the boss.

>responding to the boss' communications

Creating boss communications to occupy underlings.

>responding to everybody else's communications

Responding to communications from underlings.

>going on HR's "mandatory" courses

All this plus, ensuring actual hiring and firings are all in accordance with company policy, especially when they are not.

[yetanotheraoc]

Team standup

The good news is I won't be doing any more of the micro managing y'all have been complaining about behind my back. The bad news is meet my new AI sidekick, Micro Me.

[Anonymous Coward]

Automated meeting times!

This AI whotsit better be aware of my train timetable.

Otherwise those that start work late expect you to be available after 6pm…

[An_Old_Dog]

Re: Automated meeting times!

And this is (one place) where ML/pseudo-AI falls down. It won't realize that you being in a meeting at a client's site on one side of town, ending at 11:00, means you won't be available back at the office for another meeting, at the other side of town, at 11:15.

[ThatOne]

Re: Automated meeting times!

That's your problem, and you'd better solve it fast!...

/boss

[Giles C]

Re: Automated meeting times!

Surely then you book a meeting for the actual meeting and another one over the top of the first which is your travel time.

Or at least that is what I used to do when having to go between offices for various things…

[jmch]

Re: Automated meeting times!

Actually a bit of Google maps integration could sort that out easily...

And if that's not enough, surely the worker drones would not object to their phones relaying their location to their AI overlords in real time >>>

[yetanotheraoc]

Re: Automated meeting times!

"those that start work late expect you to be available after 6pm"

My AI can beat up your AI, and your AI knows it.

[alain williams]

Clippy on steroids

This sounds like a complete nightmare.

Then what happens when it does the wrong thing ?

These will end up being used to deal with customers, especially by those companies that try and funnel you through dealing with queries only via a web form. When your situation is not one of the top 90% and you end up corresponding with an AI ... I suspect that many will just give up (which the company will like) and, next time, order from company that still uses humans (preferably UK based).

[ThatOne]

Re: Clippy on steroids

> and, next time, order from company that still uses humans

Which you can find just next to the unicorn breeding facility. Seriously, everyone will use AI for customer contact, they already do it now ("To buy something press 1, to buy some more press 2, to be explained how to buy something press 3, press 4 to listen to the message again") so there is no reason they won't jump onto the AI bandwagon. After all it's cheaper than any outsourced call center, and almost as inefficient.

I'm thinking about a start-up to create an AI program to call those AI-driven call centers for you! Let them chat along, all night long. Instead of losing your time, the program will tell you if there was something said you might be interested about. The program needs to run on a fairly low-end computer (the kind people can spare), and ideally be able to run several threads concomitantly ("I'm calling support A since May, and support B since last summer")...

[Joe W]

Re: Clippy on steroids

There's a BOFH for that one....

[fajensen]

Re: Clippy on steroids

Well, would we be any worse off than now? Internet companies like Google, LinkedIn one cannot really contact at all, most others are using some call center somewhere with zero authority serving canned responses in poor English to not solve your problem.

Having an AI giving us the run-around would be the expected base-level denial of service, but, that AI could have a sexy voice, adapted to the user, which would be a vast improvement.

[Boris the Cockroach]

An AI

that can help me........ sounds wonderful

Ai assist: Hand me the 3/4 ring spanner

Ai assist: Now pull the oil hose out

Ai assist: pass me the new oil hose

Ai assist: run the HP pump up and open the valve slowly.

Well apart from it knowing what a 3/4 ring spanner is, I cant see it helping me out at all. although maybe..... just maybe I could use it

Ai assist: Come up with a polite insult for the dunderhead who emailed this morning.... lets do some fine tuning.. far too many fucks... lets try an honest insult... eeek we cant email that!

And finally as a serious note

Quote

"It will result in shifting more work to AWAs, and reduce overall employment levels. We will be looking at a four-day work week in five years, with growing populations of alternative non-traditional work lifestyles,"

Or going by the experience of my industry (and indeed several others) employment levels will be cut by 20% or more with the survivors having to do a full 5 day week trying to cover more ground than they did before with the rest thrown on the unemployment scrapheap.

"Gissus a job.. I can do that"

[Neil Barnes]

Re: An AI

Or going by the experience of my industry (and indeed several others) employment levels will be cut by 20% or more with the survivors having to do a full 5 day week trying to cover more ground than they did before with the rest thrown on the unemployment scrapheap.

"Gissus a job.. I can do that"

I wonder... when these companies have replaced all their staff with AI, who do they think will have money to buy their products?

[Roland6]

Re: An AI

>” when these companies have replaced all their staff with AI, who do they think will have money to buy their products?”

…Just shows universal income might be a good idea:

The government pays individuals, obviously this will need to be at a higher level than “welfare” so that people can afford “luxury” goods.

To regulate money in circulation, Companies will need to actually pay much higher taxes.

However, the low tax/small state obsessives will object, wanting people to have subsistence levels of welfare etc. and demand that the government should use its much reduced tax revenues and ability to print money to print money to subsidise their loss making businesses…

[Zolko]

Re: An AI

Just shows universal income might be a good idea

we're ready for Ark-B

[Boris the Cockroach]

Re: An AI

Quote

"I wonder... when these companies have replaced all their staff with AI, who do they think will have money to buy their products?"

They dont care, they've shown a proactive approach to reducing costs, increasing productivity and increasing profits.

Which means big fat bonuses for the C-level staff... who can then waltz off into another high paying job while their previous company crashes and burns behind them.

Like they've done previously with training staff ("No point training, its a large cost with the risk the staff go off and join someone else")

And we're seeing the fruits of that decision coming right through where companies in my line cannot recruit anyone because very few people have the skills needed.(we think the PFY only got here because someone gave her a box of lego for xmas instead of a 'my little sexist girl toy').

So I too can see the need for a basic income paid for by taxation of companies. but on the plus side for employers , if theres 20K/PA basic income for everyone, then they can reduce pay rates by that and not affect anyones take home pay. and the wage slaves can do things like take up further education knowing they'll have the basic money to survive on regardless.

[CheesyTheClown]

Re: An AI

Product cycles are an issue with cars. Any car made today, with the exception of accidents or being made by GM or Ford should last about 25 years.

That means, when automotive shops are effectively factories capable of completely and automatically disassembling a vehicle, inspecting every piece and reassembling it it will take at least 25 years for meat sack mechanics to be out of work.

And, well, somehow companies like GM and Ford will continue to exist for longer than that and they certainly will never invest in vehicle longevity. The new cars sold 25 years from now should hopefully last 50 years or more. GM can’t even make a connected vehicle that can receive remote software updates. I’m still laughing that Cruise had to recall all cars to update their software. If I ever managed a modern car project (anything from 2005 forward), the first thing I’d do is make sure every single computerized component could be fully remote managed.

[Ideasource]

Re: An AI

What a nightmare that would be.

Not for you.

But for the people who own those vehicles and others on the road.

When your security gets cracked in parallel by the ever emerging curious consequence-naive PFYs then The fallout presents real physical danger.

But assuming a harbor from liability combined with an apathy for real world consequence unto others, I agree it would make your job much easier.

[[e-mail]]

Re: An AI

Only trouble is, a lot of hands on professions will be flooded with freshly unemployed workers. Most of those workers will be fully capable of learning the job in a reasonable amount of time and effort, and wages will crater as a result.

[martinusher]

Re: An AI

They have been blathering about short work weeks and more free time for literally decades. During that time the work week in the UK has actually got longer.

Fewer jobs and more bureaucracy. Endless sales because everyone who's not part of a fortunate few will be on piece work.

Thus civilization falls....

[John Miles]

I guess someone has met some of my co-workers

as I think chatbots of last decade can be more useful than them

[elsergiovolador]

Pay

If AI is going to increase productivity, surely this means workers should get more pay?

If this is instance of doing more for less, then why would anyone bother to use the AI contraption?

That said, the way people use these LLMs is hilarious and I think it creates more friction.

Like when co-worker sends an email, they no longer feel confident to send it until ChatGPT gives the green light and edits it. Now more often they learned to write a prompt like "We need to do X, please come up with the steps and realistic time line given a two weeks' sprint. Split tasks between two developers (front end and back end) and UA. Create the meeting agenda and compose an email to send to the co-workers."

Co-workers then do something like this "I just received this email with meeting agenda. I am a front end developer. Please come up with questions I can ask during the meeting and possible answers. Find potential flaws in the timeline and risks that I can encounter."

If the meeting is synchronous. They type the questions to ChatGPT as they are asked and try to keep pace so that they can incorporate ChatGPT answer.

Comedy.

Or when you ask someone "Hey how did you do this? Can you explain?" and they answer "Why don't you just paste relevant code to ChatGPT and ask this question? Are you too lazy to even do that?"

[[e-mail]]

Re: Pay

No, the productivity quotas will increase but don't expect any extra pay. Just like some of the job ads for fluff farms hiring "writers" with the expectation of anywhere between 200 and 500 articles per week

[yoganmahew]

Re: Pay

"If AI is going to increase productivity, surely this means workers should get more pay?"

So all the mindless tasks will be automated, I'll have to think all day, more than doubling my workload. The mindless tasks let me recover and context switch. Without them I'm doing a lot more work!

[SundogUK]

Re: Pay

It's the AI increasing productivity, not you. Therefore, the AI should get the benefit.

[CRConrad]

Re: Pay

It's the AI increasing productivity, not you. Therefore, the AI should get the benefit.

Yeah, absolutely.

I mean, just like when widgets started to be manufactured on lathes in stead of whittled with a knife, part of the increased revenue was paid as wages not to the lathe operators – the workers – but to the lathes, right?

[Sven Coenye]

Re: Pay

No, to the lathe builders. Or here, the chatbot service - subscription, of course.

Greetings from Biibi Aittu

[b0llchit]

Connected plugins for the win

I foresee a new bright future and business opportunity to make boatloads of money for the AI companies. Ad-business, move aside!

The AI will connect through plugins to the cloud and all your data will be there for the grabs. The AI company will then make an all new subscription service for the competition to learn your processes. You get the information you pay for, of course. You can also subvert the use and performance of the AI for the competition, with the right subscription and the right amount of money, of course.

Indeed, a bright future I see.

[elsergiovolador]

Re: Connected plugins for the win

This will fail "can they steal your lunch" test.

By creating plugins for these AI thingies, you basically do R&D for them for free. Once your plugin becomes successful, there is nothing stopping them to do just take it (if you open source it) or to re-implement it.

Unless you have your own AI platform, creating plugins is mostly a fool's errand.

[Richard 12]

Re: Connected plugins for the win

A huge number of businesses seem to be quite happy to hand their lunch to others, and always seem surprised when it gets stolen.

[teknopaul]

Bash

Can't help but think app developers should be more diligent with cli development and bash should be taught schools.

Automation ought not to be hard enoughto require AI.

But for many people it is, because app companies

want developers to hold special magic powers instead of empowering users with trivial cli interfaces.

With a de ent cli you write your code in python or whatever if you prefer, but you can't beat the simplicity of bash for automation of cli tasks.

[mpi]

I have a question:

>In the future, instead of having to check your calendar and message back and forth with someone to settle on a time and date for a meeting, for example, Lindy's agents can connect to your calendar and email apps to automatically find a free time slot, and write and send the email asking them to meet.

It is both, supremely amusing and incredibly saddening to me, that humanity has arrived at a point where it is technologically capable of creating artificial intelligence capable of doing this ... while simultaneously still wasting an enormeous amount of time with meetings-that-should-really-be-emails.

To put this another way: Why do we use our technological prowess to automate away the pointless rituals we invented to busy ourselves, instead of doing away with the rituals, and use the compute powering our AIs to do something more productive?

[elsergiovolador]

Re: I have a question:

oh and the assumption that if someone has a window in their calendar it means it can be filled with a meeting.

it also assumes that person will enter every meeting to the calendar or perhaps enter blocks of time where they are "busy", presumably even adding a context what they are doing during that time.

We are then entering micromanagement territory.

[love not war]

Re: I have a question:

Yes. It can only work if there is a whole lot of meta data for meeting bookings.

My calendar bookings certainly don't look like that. Most are just one or two words without a location. Or a location without any other content. Etc.

And then there is the stuff that is in a different calendar. Or just not in any calendar at all.

[Anonymous Coward]

Re: I have a question:

Well AI is going to make micro management cost effective at last. No more trying to find PHBs who can handle the cognitive load of micromanagement.

[John Miles]

Re: wasting an enormeous amount of time with meetings-that-should-really-be-emails.

Now add wasting time with emails that shouldn't exist or at very least to much more focused recipient lists.

[Denarius]

Re: wasting an enormeous amount of time with meetings-that-should-really-be-emails.

Coming soon. AI driven Email storms as users AIs consult with manglement AIs who need to schedule meetings about scheduling meetings while HR AIs are inventing new courses to use all the "empty" time slots their AIs see. Makes the old email storms look mild. If any of the AIs are as useless as smart anythings I suspect the best communication process will be watercooler conferences or coffee club gatherings, leaving the IT devices to waste energy moving electrons

[mpi]

Re: wasting an enormeous amount of time with meetings-that-should-really-be-emails.

> AI driven Email storms as users AIs consult with manglement AIs who need to schedule meetings about scheduling meetings while HR AIs are inventing new courses

Very relevant video about this topic.

[doublelayer]

Re: I have a question:

Also, you don't need any AI to do this. The easy cases are already fine to handle with basic software. I need a meeting with four people in my organization, so I go to whatever calendar setup we already have, add all recipients to an invite, and pull up the shared calendar feature and look for a time where everyone says they're free, then try that and see what happens. Of course, this mostly doesn't work, but not because they weren't free when they said they were; it's sometimes the problem, but usually the problem comes earlier.

When this doesn't work, it's usually for one of a few reasons that AI won't fix. It doesn't work because everyone's in some meeting and there is no time when all five of us are free that is feasible for our use case. Or there is a time zone conflict, and trying to keep this in normal work hours for everyone is proving infeasible. If I'm the only one on this continent, it's usually fine for me to take a meeting at a weird time so that it works for everyone else, but if I have a colleague in the same position, I would probably not do that without apologizing to them and getting their permission to do something like that, which an email probably won't do. Or it doesn't work because one of the four doesn't want random meetings scheduled, so they've blocked out multi-hour chunks as busy time when they are working but could be available, but I won't know that without asking them. How hard I have to ask them depends on the importance of this meeting and their presence at it, and their reaction will depend on the same thing. If it's me organizing it, it's probably pretty important because I hate meetings, which also means my colleagues are usually happier to attend something I've organized, but the AI won't know that, nor would their responding AIs.

This all gets more broken when I drop the "in my organization" part of the premise. The AI doesn't have access to the calendars of anyone outside the company, and if I have to meet with them, then we're dealing with a lot of possible ways to ask for and get scheduling information. The AI could probably deal with some of this, although I'm worried that it will get confused about someone's time zone data and make a complete mess when I'm not looking. There's a reason there are at least two companies I know of who offer a calendar reservation application as a product. An AI that lacks information will do it badly and giving the AI sufficient information to manage it is a sisyphean challenge, both in the "a lot more work than is justified" perspective and the fact that it will never end.

[Doctor Syntax]

Re: I have a question:

This all gets more broken when I drop the "in my organization" part of the premise.

Again this doesn't need AI to do that. All it needs is a new internet format and protocol for exchanging calendar information so that ordinary rules-based software can try to find an optimal solution.

What's that? .ics files and WebDAV?

Nothing new under the sun. And that includes solutions searching for a problem, even an already solved problem.

[doublelayer]

Re: I have a question:

Of course you can exchange calendar data with other organizations, but I don't and never will. My work calendar has data they can't have, and it's not like I can easily grant or revoke access to certain parts of it. This means that, unless we need to meet so frequently and irregularly that it becomes a necessity. Otherwise, I don't want people outside the organization to know who I'm meeting, the titles of the meetings, whether I'm rejecting their invites because of a real clash or not. Similarly, I wouldn't want my employers to be able to tell if I'm having an external non-work related event by seeing who has accessed my work availability. That's why less invasive methods are used.

[yetanotheraoc]

Re: I have a question:

`...giving the AI sufficient information to manage it is a sisyphean challenge, both in the "a lot more work than is justified" perspective and the fact that it will never end.`

From the point of view of those building the AI, this is a solved problem. Just slurp everything. You will agree(*), won't you? For the sake of a little convenience?

(*) I'm well aware you already said you would never agree, nor would I. Just pointing out that the future data needed for the AI to function as marketed IS PART OF THE PLAN. In order for the AI to schedule, it needs all your calendar all the time, as well as all the calendars of all your contacts all the time. It only hurts at first and after that you will start to like it.

[fajensen]

Re: I have a question:

Why do we use our technological prowess to automate away the pointless rituals we invented to busy ourselves, instead of doing away with the rituals, and use the compute powering our AIs to do something more productive?

Because, the point is, that the rituals must be performed. Indeed, any large organisation can be said to exist primarily to provide the funding and set the schene for the performance of the rituals.

The problem we have with people is that they can manifest things. Humans, left all on their own, they will manifest distracting, maybe dangerous, even terrible things. Like the Golden Jesus* running for President and then Righfully slaying all the Impure and Improper starting with the gays. So, a network of distractions is created, keeping the human mind busy with insignificant trivia and white noise so Bad Things does not happen too much.

This network of distractions we call "Real Life (tm)" or "A Career". It got damaged during Covid 19 and we lost containment somewhat, allowing lots of people too much freedom to manifest their inner nuttines - and set their creations lose on the world. This is obviously not good so ... everything will be wound back and tightened down harder, with better distractions and more elaborate rituals to perform. AI will be a critical part of that work.

I.O.W. We will have less free time, more distractions, and more performances in the future!

*) I just see a fat-ass fuckhead loser, but, the phenomen no different from the happenings in Project Blue Book, where a bunch of people see a flying saucer with 3 beings in it waving at them, while the radar sensors and the Air Force sees a rocket stage de-orbiting.

[Anonymous Coward]

tedious administrative chores that suck up people's time

ah yes, I think I have heard it before, more than once. Something about those devil-powered 'steam' machinery, then, I'm sure, somebody must have used that argument about typewriters. Not to mention computers and the internets. By now, I should be practically work-free with massive amount of time on my hands. Oh, wait, this rings a bell, where's my last unemployment benefit payment gone?!

[Snowy]

Sure

All going to happen soon™, compared to the heat death of the universe what is a few decades.

[ThatOne]

AI knows better what you need

> Large language models will evolve from AI chatbots generating synthetic content on your screen to virtual agents that are capable of performing actions on your computer.

Let me rephrase that in a less idealized and more realistic way:

"Large language models will evolve from AI chatbots generating useless chatter on your screen to virtual agents capable of selling you stuff you neither need nor want."

Obviously AIs will get monetized, i.e. they will be trained to try to sell you stuff. Now give them decision powers (somebody like Microsoft will rush to do it) and they don't even need to convince you anymore: You just pay the bills and collect the parcels. The worst is that a lot of people will be quite happy about that ("shopping is such a chore... I never know what to chose... The AI cares for me, I get a gift each day...").

[bemusedHorseman]

More like...

"Your coworkers have become dumb enough that AI chatbots can replace them and no one will notice"

[Jlong332]

Chatbot creating another chatbot

[Jlong332]

I don`t want to get up. My chatbot will work for me today.

[yetanotheraoc]

Even better

Your chatbot can call in for you. After a while the checks will go to your chatbot, too.

[Dacarlo]

Re: Even better

What are they checking for?

[Starace]

Reinventing the wheel with AI

Lots of these ideas sound just like stuff that either already exists, or existed then died from indifference, just with a lot of expensive AI being used this time to try to make it work.

I guess maybe this time the difference will be it convincingly looking like it's doing the right thing while just inventing the details from thin air instead of using what you actually wanted.

[Tron]

The next big thing in tech...

...will be reliable ways to turn all this AI BS off. Just as we find ways to disable all the new, unwanted gimmicks MS put into new Windows versions.

Ideally Glorious Leaders would mandate an AI off button. But as they are siding with AgriPharma forcing us to eat GM food by banning labelling, they will no doubt allow Big Tech to force this crap on us too.

We really should elect less crappy governments.

[trevorde]

Things AI will never be able to do

* heat up fish in the microwave

* knick your parking spot

* forget to flush the toilet

* take the last chocolate biscuit

* borrow your cup

* have an affair with the boss

[sten2012]

Re: Things AI will never be able to do

I absolutely don't believe the last one.

And in fact said boss could probably then be convinced by the AI to do the others:

"Honeyboss, I notice you are 15% more efficient at your management duties when you have your omega 3. Run out and get a kipper, oh and I think there are no plates clean so grab the cup from cubicle 6 on your way through and microwave it in that. There's a parking spot on bay 7 closer to the building - that should save you time"

[Dacarlo]

Re: Things AI will never be able to do

"* have an affair with the boss"

Just wait till robotics, AI and RealDoll (tm) team up. There's no problem that IT can't make more interesting.

[Bebu]

Re: Things AI will never be able to do

《Just wait till robotics, AI and RealDoll (tm) team up. There's no problem that IT can't make more interesting》

Brings the thread back to the first post re wives/ spouses :)

I wouldn't invest a single razoo in LLM/AI but the AI driven soft androids (or is that gyndroids?) could be the killer app as it were.

Some of the more... ah interesting parts of the internet should make an "interesting" training set.

"What are watching?" "Supervising the bot's training, boss." (Nice work...)

What is a bit scary is that most of the technology is already in place and a functional artifact would cost a lot less than six million dollars.

I recall some prurient speculation as a young teenager amongst my peers as to what parts of "The Bionic Woman" were bionic.

I don't suppose Richard Branson would be interested in bankrolling a startup enterprise offering as Terry Pratchett put it "negotiated affection" services? - He owns the most apt brand for it as that part would definitely be user replaceable.

[Anonymous Coward]

I might have found a use for a meeting AI...

...as score keeper and umpire for meeting wankword bingo games.

[yetanotheraoc]

Re: I might have found a use for a meeting AI...

AI wins again!

[Omnipresent]

Boss leans back in leather easy chair and lights a cigar.

"Hey Windows!"

...ding, ding, dong.... "Yes Mr. Richard?"

"Locate Jones, P"

...Queuing request, queuing request, queuing request....

"Jones', P last retinal scan was used at 11:13 hrs. to enter the employee toilet."

"Send his AI watch his pink slip."

"Jones, P does get a fifty nine minute lunch break Mr. Richard, are you sure you would like to send him his separation papers?"

"Fine, Cut Jones' P. Lunch break to 29 minutes instead, and cut off his retinal access to the privy."

... ding, ding, dong.... "Windows has cut Jones' P lunch break, and also his retinal access to the employee bathroom. Would you like windows to scan his AI for more information on Jones, P?"

[gweedo]

Ridiculous timeline. LLM will be helpful for knowledge workers for those big businesses willing to pay the high price for adding these features. I've used many of the co-pilot solutions available today, and they are for sure helpful, but in no way are they actual AI. There is no intelligence or thinking going on, they are just really good uses for the LLM to summarize, reconstruct words & materials, etc. They all need someone knowledgeable and time/effort to get something of value from them. They are great at getting something started, but you still have to bring it all together into something that is useful. AI that everyone thinks of when raving away on this subject are decades (at least) away. That doesn't take away from the usefulness of the LLM tools coming.

[ecofeco]

Eventually?

It's already smarter than most of my co-workers.

[yetanotheraoc]

Who's the chatbot now?

"... paralegals will take on more client relationships and advice support assisted by agents that have passed the bar exam and provide legal services at a lower cost than a licensed attorney."

So the paralegal is now a presenter reading from a teleprompter run by a licensed AI. No ad-libbing, just stick to the script, please. And do try to smile, your next break is in 15 minutes.

[cookieMonster]

Our Northstar is that …..

The wank is strong in this one

[Ian Johnston]

Much the same bullshit hype as we were bombarded with about blockchain. Yes, if your job includes stuff which can be done by a glorified autocomplete, it can probably be automated. But then again, you probably needn't have been doing it in the first place.

[HuBo]

Rosy mixed-metaphoria in the bumble salesdroid

Sure is good news to hear that the painful dread of backbreaking daily toils may be greatly eased by newfangled magical thingamabobs, or even replaced entirely with extensive leisure time to spend at city museums and live art performances. Just as long as it works as advertised on the tin, and doesn't covertly rob us of our hard-earned means of sustenance, in this dog-bites-man-bites-dog lowest plane of the great Ponzi pyramid of wealth concentration, and its eponymous giant sucking sound.

To this effect, and within the context of its recently inaugurated partnership with Microsoft, the AFL-CIO could do worse than to mandate that these revolutionary new means of production, say genAI, be owned by the workers currently performing the tasks to be eased by these emerging tools. Own your AI (grow it locally, like a victory garden!).

[[e-mail]]

AI doesn't need to destroy humanity.

The humanity will merrily self destruct when unemployment rates hit 20%+ and most of those workers have zero options apart from terrible service and mental labor jobs and a lifetime of poverty wages.

COVID was a great preview of what happens when a sizeable part of the population is put out of work overnight. A lot of governments stepped up with wages replacement, only because the alternative would be a lot more expensive.

Some of the folks in AI circles understand this, but the other half seem to think that getting shot for their wallet and shoes on the way back from work can't possibly happen to them.

[Bebu]

Re: AI doesn't need to destroy humanity.

"humanity will merrily self destruct"

Modern humans will likely survive as a species even when confronted with most conceivable catastrophes where survival is actually possible.

What is looking more and more likely is the destruction of civilization in a lot less than a millenium. Slowly at first, slightly faster then abruptly whack ... gone. Could be back to banging rocks together in fewer than 300 years. When systems fail its rarely just one factor or even a handful - its a large number of interrelated factors in what was a relatively stable steady state that suffered persistent disturbance(s) that ultimately leads to instability and potentially collapse.

[Jilara]

Meetings are already automated

I work for Big Tech, and before this AI nonsense, our meeting calendar could already find free banks of time for multiple people, plus add an open conference room, and send out notifications with just a few clicks. But it allows choosing options, like getting to pick the timeslot that isn't hard up against someone's commute bus time. But it also flubs conference rooms and occasionally double-books. But this kind of functionality isn't rocket science. Makes me think of when (back in the dark ages) I found a Z-80 chip being used as two AND gates...

[Michael Wojcik]

In the future...

... we'll have AI agents to save us from such time-wasting activities as communicating with other people (heaven forbid we have to discuss when we'd like to meet!) or thinking about how to describe our activities.

Terrific. I feel we're really not plunging into learned helplessness and intellectual laxity as quickly as we might.

I've run across some actual examples of actual utility from LLMs and other applications of large transformer models, such as the recent paper from Wong et al. explaining how they used an explicable LLM-based model to discover a novel structural class of antibiotics. If that work pans out, great. However, the vast majority of everyday use of LLMs I've seen has negative utility: it's just a way to avoid something which would entail learning or improving a useful skill, developing a deeper understanding, or at least serve as an intellectual paideia.

[HuBo]

Re: In the future...

Superb comment! I hope you won't mind if others re-use your phrase "intellectual laxity" in other contexts, for example, in academia:

This homework of yours! It resembles the unfortunate byproduct of a game of intellectual laxative roulette! Zero ... flush ... and well-deserved!

(P.S. nice to see the ref to "explainable, substructure-based [LLM-oriented] approach" to novel antibiotics!)

[CRConrad]

Laxity...

...is not quite the same as laxative.

[captain veg]

Ideally, it would even add a Zoom link or Google Maps directions to a place too.

This must be some new meaning of the word "ideally" that I was previously unaware of.

-A.

[steelpillow]

Technical author here

We already have the things taking over chat helplines and "help" websites, how long before my job is reduced to an editorial one - pointing the AI hack at the design documentation, no doubt churned out by another AI under the aegis of the SDA, and proofreading the outcome? I suspect that an attractive SDA is going to be the key decider in my next career move, we'll have to find something to do while the Microsoft azurey-cloudy-sharepointy-too-many-tentacles nameless horror reboots yet again - and again - and....

[Grunchy]

HAH

I worked with some ACTUAL intelligent assistants, and they were a complete disaster.

The circumstance was that we had a mountain of drawings to be updated per redline, but the majority of them were “simple” updates. So the boss had a great idea, let’s hire some intelligent assistants in India. The workflow goes: bundle up all the simple, busy, impossible to get wrong work (into a zip file), email that to India, and spend all the rest of your time doing the more specialized, challenging stuff. Then tomorrow morning, briefly check all the work from India, and release all the updates at once.

But what really happened: everything sent to India was mistaken, or poorly done, or introduced new mistakes. It’s as if you gave your information to hackers to mess it all up, and then they billed you for the privilege.

The amazing thing is that this persisted for months! Well it wasn’t hard to figure out. The boss got a raise for “solving” the problem, and he had a “closed ears” policy about the solution not working. Plus even better, he made it so that all the performance metrics were produced exclusively by the Indian vendor with no input from the on-site design and drafting department (as a matter of fact — we were deliberately gagged). Needless to say, our outsource colleagues published an unbeatable 100% on-time, 100% error-free performance record, justifying several rounds of bonuses: all fraudulent.

How did the whistle finally get blown? The dept VP finally came for an independent tour and asked to see the “success” with our online friends. Boy was he surprised to learn it actually took more time and effort to correct the “intelligent helper’s” edits than if we were to simply do the original job without help. Our actual workload ballooned from 100% to something more like 140%.

My prediction: AI helper bots prove to be a far greater hindrance than net useful benefit for all the foreseeable future.

Exactly like “full self driving.”

(Trust me: come test your chauffeur robot in a Calgary black-ice blizzard, to truly understand the meaning of “futile” !)

[Anonymous Coward]

Will we work less or more?

This should have been phrased 'Will we work more for less?'. Without the question mark.

[Snowy]

Smart enough to replace your co-worker.

Then you can train it up to replace you too!

[Anonymous Coward]

An automated co-worker? Probably not...

Lindy, reboot machine xxxxx. (Yes, it could probably do that.)

Lindy, go replace the thin client in room xxxx. (Nope, no arms.)

Lindy, do a complete reinstall of system xxxx from the CDs in the safe. (Nope, no arms, no fingers, instructions to complex.)

Lindy, figure out why xxxxx is happening when the process gets to step yyyyy. (Not a chance.)

I'm not worried about an "AI"* replacing or being a co-worker. It would be utterly useless!

* They're rather heavy on the A, and entirely empty on the I - these are language models, less intelligent than your average cow.

[Rich 11]

Not holding my breath

... in the short term, one to three years from now, autonomous workplace assistants (AWAs) will be able automate away easy tasks that take no longer than a few minutes for a human to perform.

I'm pretty sure I was first promised this 30 years ago.

[Paul Hovnanian]

Wally,

... to my Dilbert.

[Blackjack]

[ smart enough to be your coworker] is a very low bar to clear. Not even joking, Ihave seen people so stupid I couldn't figure out how they keep theit jobs besides nepotism.

[Dacarlo]

Chilling

"I think it's people who've been stealing robots' jobs..."

Never have I heard more chilling words spoken.

[Ken Moorhouse]

Electricity

This is the elephant in the room.

The amount of power AI uses is going to be horrendous, but because it is not a direct cost nobody will realise it until it is too late.

Too late? You know these conversations everyone is having about fossil fuels these days? Well, it's a similar thing.

[Blackjack]

Re: Electricity

Actually they will realise it once the power bill comes knocking.

[Anonymous Coward]

Memorize you passwords for you, and enter them without being asked

pure genui</ssssssssssssssssssssssssss>:(

[tyrfing]

The largest problem will be data.

An AI can only schedule two people who are willing to share their entire schedule with the AI. It can only order things if you tell it what you would like, and allow purchases on your card. Filling in invoices allows the AI an enormous amount of inside information on your company.

But people are becoming unwilling to trust companies with this information. Data mining for it just makes them even more unwilling, and shows you really can't trust them.

It's not possible to run the leading edge models on personally owned hardware, and companies are unwilling to release those anyway. So you can't prevent unauthorized usage that way.

Think of putting a person in the place of that AI - except the person is controlled and paid for by someone else. Do you trust that person?