Job seekers call BS on the workplace AI revolution
- Reference: 1732129869
- News link: https://www.theregister.co.uk/2024/11/20/resume_genius_ai_survey/
- Source link:
As part of a [1]larger study into the state of job hunting in 2024, CV pushing platform Resume Genius looked at attitudes toward AI among 1,000 US job seekers and [2]found that 69 percent doubt AI's ability to boost their work performance, while 62 percent lack faith in AI's capacity to reduce their workload.
Consistent with the majority opinion that AI in the workplace has failed to impress, only 34 percent of respondents said they were worried about being replaced by a bot, while just 30 percent think AI will increase competition for jobs or harm salaries.
[3]
Broken down by generation (Boomers, Gen X, Millennials, and Gen Z job seekers all responded), the results are largely the same, with even Gen Z workers skeptical of the latest "next big thing" in enterprise tech.
[4]
[5]
In short, Resume Genius's findings align with other recent studies suggesting enterprise AI's hype has not lived up to its marketing promises.
Nobody wants this except sales teams
One need not go far to find copious evidence of the lack of enthusiasm among workers for enterprise AI – we've covered it plenty on The Register .
Back in July, we covered a study into AI usage in the workplace by Upwork, which found that 77 percent of employees pushed to use AI productivity tools at work have [6]ended up less productive . Many employees report feeling pressured by management to use AI tools into which their employer has sunk money, with 65 percent saying they're struggling to satisfy increased demands that bosses expect AI to help them meet.
[7]Forcing AI on developers is a bad idea that is going to happen
[8]How deliciously binary: AI has yet to pay off – or is transforming business
[9]Microsoft unleashes autonomous Copilot AI agents in public preview
[10]If you use AI to teach you how to code, remember you still need to think for yourself
That's not to say workers wouldn't be willing to embrace AI, with Adobe (itself marketing an enterprise AI tool) [11]finding in a September study that 80 percent of workers would do so if it could actually save them time.
Of course, reality appears to differ when the rubber meets the road, with AI data services firm Appen releasing a report last month that found the [12]return on investment (ROI) for AI projects has declined and few investments reach deployment.
While Appen attributes much of this to a lack of good training data, Gartner [13]suggested in May that "difficulty in estimating and demonstrating the value of AI projects" was the primary obstacle to successful AI adoption.
[14]
In other words, while the C-suite is won over by glossy presentations and hype from companies like Microsoft, with its " [15]Copilot everywhere " attitude, it's not paying off in actual usage.
Whether AI vendors will take notice of such shortcomings is doubtful. Microsoft certainly hasn't. Redmond [16]increased prices on Copilot by 5 percent when introducing monthly billing options last week that still include a year commitment. Meanwhile, employees seem unsatisfied with the tools being pushed on them. Hopefully things will get better before [17]AI makes them worse . ®
Get our [18]Tech Resources
[1] https://resumegenius.com/blog/job-hunting/job-seeker-insights-survey
[2] https://resumegenius.com/blog/job-hunting/job-seeker-insights-survey#ai
[3] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=2&c=2Zz5qECqfLBQIO550D_-5AQAAAQ4&t=ct%3Dns%26unitnum%3D2%26raptor%3Dcondor%26pos%3Dtop%26test%3D0
[4] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44Zz5qECqfLBQIO550D_-5AQAAAQ4&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[5] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33Zz5qECqfLBQIO550D_-5AQAAAQ4&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[6] https://www.theregister.com/2024/07/26/ai_hinders_productivity/
[7] https://www.theregister.com/2024/02/12/opinion_column_on_forcing_ai_features_on_developers/
[8] https://www.theregister.com/2024/07/30/ai_has_yet_to_pay/
[9] https://www.theregister.com/2024/11/19/microsoft_autonomous_copilot_ai/
[10] https://www.theregister.com/2024/01/27/ai_coding_automatic/
[11] https://www.theregister.com/2024/09/05/aishy_office_workers_ignore_concerns/
[12] https://www.theregister.com/2024/10/22/genai_roi_appen/
[13] https://www.gartner.com/en/newsroom/press-releases/2024-05-07-gartner-survey-finds-generative-ai-is-now-the-most-frequently-deployed-ai-solution-in-organizations
[14] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44Zz5qECqfLBQIO550D_-5AQAAAQ4&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[15] https://www.theregister.com/2023/05/10/microsoft_copilot_ai/
[16] https://www.theregister.com/2024/11/14/microsoft_365_copilot_monthly/
[17] https://www.theregister.com/2024/11/18/opinion_piece_ai_tools/
[18] https://whitepapers.theregister.com/
It will happily invent non-existing packages and API calls, it sometimes generates API calls that are deprecated, don't take the right arguments, etc.
In essence it's a very weak oracle function that can save time in situations where you can verify the solution quickly (<1min), and the time it would take you to do it by hand would be 1-2 hours or more.
It is somewhat better than Google at finding probable candidates for API calls and commands where documentation is outdated/absent, but this is also due to Google getting worse over time, not so much AI getting better, or is that just me?
Overall office life could improve if managers were to test and use the tools they force on the workforce. But on to other dreams.
In technical, hard fact, fields AI-LLMs are useful because when they are not useful they are constantly and immediately corrected by the human likes of you. You learn what requests are too general, so you will refine your query it down to a level that the AI-LLM can handle. Their power comes from your power.
This survey is about AI in the greater and wider job market. To give you an idea, here is a quote found in a recent Reg article: "Microsoft unleashes autonomous Copilot AI agents in public preview" -- "Agents built in Copilot Studio can operate independently, dynamically planning and learning from processes, adapting to changing conditions, and making decisions without the need for constant human intervention," explained Charles Lamanna 's Copilot Autonomous Agent , corporate VP of business and industry for Copilot, in a blog post. "These autonomous agents can be triggered by data changes, events, and other background tasks – and not just through chat!"
Imagine being a recent Liberal Arts graduate and being told to implement "that", being totally confused, and then soon find your job outsourced to someone overseas who will fail equally hard for 1/8 of the cost.
I've been saying this for a while. You can lead an AI (Not AI but a chatbot) to water but you can't make it understand why it needs to drink it.
Sure it will get you going but the way I look at it is that you can tell it to make a thing but you can't tell it to think about the thing it is making. It will take zero consideration to anything else. Security? Nah, Optimisation? Nah. Will it do what you want? Sure but at what cost?
You may think it takes the grunt work out of it but at what point do we or others no longer understand the grunt work? When does a developer no longer understand sanitising inputs? Chatbot doesn't care about that and never will. Those easy peasy tutorials to get you going are it's bread and butter.
I'm probably waffling here but I think my point is made.
Time savings
"study that 80 percent of workers would embrace AI if it could actually save them time."
Boss: "You are saving so much time, you can also fit X, Y and Z into your day now."
Re: Time savings
Boss thinks a bit more.
"If you are using A.I. to do most of the work, why I am employing you? You're fired."
Re: Time savings
I would prefer the second boss over the first. More short term pain for me but I know he'll get his comeuppance. First boss will just keep screwing me over until I quit and then he'll just carry on the rinse / repeat cycle on my successor.
The emperor's new clothes?
See title...
Yeah, it saves no time overall if you care if it's correct
There are two specific use cases for me where it actually saves time:
- As code autocomplete of less than 4 lines - and it has to be integrated with the client too, just tab to accept, or it takes more time to write the query than it does to just write the damn code myself. Of course it also regularly suggests totally wrong code which you have to hit a key to bail on, so not sure it saves all that much time.
- When I need a graphic generated because my art skills suck.
Other than that whatever it does takes so much checking and fixing I can just do it myself faster. Why would I need it to read and summarize a document when I have to read it enough to make sure it's not bullshitting anyhow? And then I can just summarize it myself. Of course there are a lot of people who are just YOLO copy and paste whatever it spits out and for them I guess it saves a lot of time at the expense of being non-zero levels of wrong.
It's also fast for translating from a language I don't know to English... but then it's much safer to use Google Translate because that doesn't hallucinate (it just gets confused, which is easier to detect, those manifest differently).
So I see no vast time savings to be had anywhere except art generation unless you just embrace it being wrong 10-90% of the time and DGAF.
I use GitHub copilot a lot at work and I do find it useful. It’s good for drudge work like documenting code and making mock data for unit tests.
However, you do have to be careful. I got it to write a whole suite of unit tests for some old code that was missing coverage. It generated a ton of test code in a couple of minutes which seemed great.
I then had to spend an hour fixing all the hallucinations! It saved time overall but not as much it first appeared. It’s a useful tool but just a moderate benefit and not yet the game changer it is billed as