So what jobs are AI proof then?

[StrongJamie]

Based on recent media report we are facing an imminent (?) AI revolution.

I imagine that doctors, teachers, lecturers, accountants and civil servants can easily be replaced by AI and tailored professional AI software. I am guessing that jobs that require intricate physical handling are less at risk as it would be expense to mass produce the Hardware. Hospital doctors are more at risk than ward nurses but less than GPs who soon will be obsolete.

It will be a good while until they can mass produce robots that do humans jobs, which require a lot of running around and haptic skill but some jobs don't need a person, they just need the right software (e.g. GPs).

I imagine it like this, you log onto your GP AI service, they know all your medical history and also all the up to date epidemiological data of your neighbourhood as well as your biomarkers, pulse, heart rate etc uploaded via your smart watch continuously. The system knows about all possible diseases and conditions and based on your biomarkers and symptoms knows how to signpost you for further tests or what to prescribe. Job done, no more GP.

Teachers? No need. Ai robots, virtual or physical deliver synchronous teaching the rest is done online.

Please pick holes in my assumptions or add to the list of soon to be obsolete professions.

Which ones are ai bullet proof?

You are working on an outdated version though.

I hope so. I so don't want to do this task.

Standardising dates in a spreadsheet shouldn't need an AI. ChatGPT type things certainly aren't the right tool for every task.

To all those saying that teachers can't be replaced, I think you need to reconsider. Whilst I agree that young children need to be managed, the teaching element of learning to read and use basic maths is already available online.

AI will change the way the world works and current models of teaching are already outdated. There will be absolutely no reason for pupils to learn to regurtitate facts in exams - it is already a pointless exercise when anyone can find out whatever information they need on the Internet.

AI will (and should) change secondary teaching education beyond recognition as whole exam syllabuses will need to be changed. Pupils will need to be taught to understand concepts and learn how to manage AI in particular fields.

AI can now write (basic) university standard essays on any topic. We need to change the mindset and rather than fight against it by banning its use (presumably Turnitin will be followed by something that checks for AI), we should allow students to embrace it and see where it takes us. Taking the drudge out of learning will free up time for more creativity and development in all fields.

Critical thinking will become much more important as pupils are taught to navigate the vast array of information available to them. This is a skill sadly lacking in a large percentage of the population and with growing reliance on the Internet, is becoming a serious issue - we have all seen the disasterous results of the lack of critical thinking by both the UK and the US electorate!

Guessing you used GPT-3.5. Look at what Copilot and Jarvis are doing using the GPT-4. This is already a solved problem.

And if it isn’t, there will be a tenfold increase in compute in 18 months which is already signed off with NVIDIA. They just need the H100s to be installed and training started.

@AlaskaThunderfuckHiiiiiiiii These are mutually exclusive things that you’re assuming legislators are fully aware of, despite them being on the back foot on basic social media impacts. If you think the retirement age plans of decades past are related to developments that have only happened in the last five years or so in deep learning, well, it just ain’t so.

People were still breeding masses of horses and worrying about manure piling up in streets, even after the automobile was first displayed publicly and seen to be a curiosity

Also, as mentioned before, a lot of problems from getting adequate solutions to questions or commands is in good prompting. If you tell someone “make me rich”, that can lead to massively varying outcomes, if any. But if you ask “how can I invest these savings in a business I can work with. Here are some things I have knowledge of and a passion for”, then you narrow down possible paths.

Because of the lack of long term memory and recursive learning, the model is horribly limited for safety purposes and to reduce on compute time. We’ve already seen Reflexion and AutoGPT, to name just two, that have improved on the publicly available system. We know that the latest model from OpenAI was held back more because of papers querying it prior to the alignment training that hampered certain cognitive functions.

Five years. We are now likely to have a true AGI in that timeline just off the current system designs being optimised. Factor in hardware improvements already in the pipe, and you have something pretty terrifyingly rapid in how fast it’s coming at us. And no one is really preparing.

I say this as someone who was reading Superintelligence in 2014 and seeing negligible real movement on expert systems since the ‘90s.

I work in heritage and I think there are elements of my job that are pretty AI proof.

I can see AI being used for appraising and cataloguing digital deposits. But where I work we have nearly ten miles worth of physical records spanning 800+ years. Unless there's a huge investment in digitising, i can't see archive or indeed conservator jobs being under threat anytime soon. Record managers would also be fine.

It's translating a date narrative into a standardised date format. SO there's text in there. Multiple dates, different types of dates, different languages. Think like a catalogue written by a traditional librarian at different points in time not following a standard. It is beyond Excel's own capabilities. My developers didn't do that well at sorting it out last year (they didn't have a lot of time to be honest). It's a bit complex but really not that difficult. The problem was there were maybe 10 instructions that needed to be followed but Chat GPT didn't seem to be able to do them all. And didn't manage to do more than 20 lines.
I'd like to know what tool I should be usinh.
And if Chat GPT can't follow 10 simple instructions across 100-200 lines, what can it do?

If you were using the free version, then you're running a model that isn't even the version Microsoft's early Sydney chatbot was using. If you want something specialised in using the APIs from Office to do such data collating, then that's where Copilot comes in, which will be rolling out in the coming months.

You can't really get an LLM to do everything without access to computational tools that act as plugins to enable functionality that isn't the irreducibly complex language modelling it was trained on. LLMs suck at maths, so that's why you can now have Wolfram Alpha connect to GPT-4 and utilise Wolfram Language to do much more efficient data management.

It's like getting a linguist who can write good prose to do Fields medal maths without a calculator. There's just a better way using plugins.

@Liebig Thanks for your help. We do have access for a paid for version at work. I'll find someone more qualified than me to check it out. I only had a go to what it could do as it seemed like such a do able task.

Capabilities aside, I was a slightly horrified when I had to stop it because it was showing me someone else's data (some kind of academic timetable. Definitely not mine!)

It was also frustrating that it managed to do it (almost) right once. but then couldn't repeat the process.

I'm no expert, but my take is a lot of professional jobs will simply change. AI is a tool, and the skill be how to use it and integrate it into your workflow.

In my field, engineering, I use Excel a lot. 50 years ago, if I wanted to sum a column, I'd have to enter them in a basic calculator. 100 years ago, I'd have to get out pen and paper. Now, I do SUM(A1:A10) and I get the result. The point is I had to learn how to use Excel.

In the same time, I had to learn how to set up a computer model - I didn't write the code and I don't crunch the numbers, but I had to learn how to set up the model and decide what tests to run, and how to interpret the results, and then present it to the client.

Maybe "human calculators" were replaced, but the engineer who decided what calculation to perform is still there. I suspect the same will be true of AI.

Take programming, you might not have to write a lot of the basic code ("write me a function that enters a list, and spits out something else"), but there will be people who have to decide what the client wants and how that could be provided and at what cost.

I don't use excel, whenever I've tried it seems a bit shit and clunky. But then, I've got a nice data pipelining system that we use to write complicated scientific applications (which nowadays may well include some machine learning), I reckon it could read a spreadsheet and it'd be a few simple lines of script to standardise dates and write it back out. Easy problem if you've got the right tools.

@mellicauli Yeah, that’s one of the issues that is at the forefront of this AI debate: safety, not just in terms of Skynet, but from otherwise banal stuff like if you give this system open access, what happens to privacy? If it’s trained on data scraped from the whole web using web crawlers, does that include things we shouldn’t allow it to utilise? We already had the hallucination that led to a defamation case being levelled at OpenAI by someone being accused of being something they’re not. Controlling for what data is allowed (despite safety training, some have managed to get ChatGPT to go in depth on chemical weaponry manufacture or scam ideas, for instance) is a tricky one up there with general alignment. And in an age of social media, there’s plenty of data to abuse.

@PapitoSpice So long as we haven’t all been literally eradicated in some way, an AGI basically just needs direction from a human on achieving a goal. It’s still going to need a human to justify its existence, just as war will always have humans in the loop, even if it’s drones and other robots fighting in a battlefield. You can’t have AI art without humans to appreciate it.

That said, we don’t know too much about consciousness. Who’s to say the other emergent capabilities to come from larger models won’t include a level of sapience and will?

Does anyone know what Nigerians mean if you ask what their occupation is, if their answer is "in business" ? People in UK or Brazil don't routinely just say "in business". But they do in Nigeria. I need to figure out how to match up this occupation answer in Nigeria to compare with distribution of Brazilian answers (where nobody said they were "in business").

Until AI can figure out that kind of problem, and a dozen others like it, just to get one tiny project done, my job may not be threatened by AI.

That isn’t an occupation. It’s a description, and a vague one at that, since no one can infer from it what it means.

I think psychologists are safe?

Actually that's funny I was just remembering my sister coming home from school in the 80s and telling me about a program called Eliza which was an online pyschologist0 chatbot. I think they (literally) dialled into the local university network to access it. So it's all been a long time coming really. Not the overnight explosion everyone is talking about.

en.wikipedia.org/wiki/ELIZA

I asked the question to ChatGPT, and this is what it said:

It's difficult to predict with certainty which jobs will never be replaced by AI, as technological advancements and the capabilities of AI are constantly evolving. However, here are some examples of jobs that may be less likely to be fully replaced by AI:

1. Jobs requiring high levels of creativity and imagination, such as artists, musicians, and writers.
2. Jobs that require high levels of social and emotional intelligence, such as therapists, social workers, and teachers.
3. Jobs that require physical dexterity and mobility, such as plumbers, electricians, and chefs.
4. Jobs that require complex problem-solving and decision-making in unpredictable situations, such as emergency responders, pilots, and judges.
5. Jobs that require a deep understanding of human behavior and interactions, such as anthropologists, historians, and psychologists.

It's worth noting that while AI may not be able to fully replace these jobs, it may still play a role in enhancing and augmenting them. Additionally, there may be new jobs created as a result of advancements in AI, such as those related to the development and maintenance of AI systems.

It's funny, I think my job could be done by AI if the customers were competent enough to answer the questions correctly but I ALWAYS have to ring them to check what the nonsense they put down means.

I look forward to being offered an incentivised inducement to retire (civil service, who knows)