Mon Sep 18, 2023, 01:12 PM

Bad news, star employees: You're not the ones who'll benefit the most from AI (Business Insider)

https://www.businessinsider.com/ai-employees-lower-performers-most-impact-boston-consulting-group-2023-9

-snip-

Before the experiment, the staffers were tested to gauge their average level of performance, classifying them into "bottom-half" and "top-half" skilled participants.

-snip-

The greatest gains were seen by below-average performers using AI, whose average performance improved by 43%.

Their above-average counterparts only saw an average performance increase of 17% from using AI.

-snip-

Consultants with access to AI performed up to 20% worse when presented with tasks that were beyond the AI's grasp. In these cases, the AI would present misleading yet plausible responses.

-snip-

The article mentions another study done in April, where a Fortune 500 company tested its customer service workers. In that test, the most highly skilled workers got ZERO increase in productivity from AI, while the least skilled got a 35% increase with help from AI.

The same may actually have been true with this new study, that the very best workers in the new study had zero increase in performance using AI, but that was averaged out with those toward the bottom of that top half of workers.

What this means, probably, is that a lot of employers using AI will choose to lay off all or most of their most highly skilled workers, who are probably also their most highly paid workers, and replace them with new workers paid much less.

I heard something like that from a CEO of a company on a panel about AI, a video of which I posted here months ago. He said in the future he'll probably just hire coders for his company straight out of high school. The degrees in computer science won't be needed, and the kids will work for less, but still be able to code with the help of AI.

Of course, there is that not-so-little problem of the mistakes made by the AI itself, as mentioned in the last paragraph of the excerpt I quoted.

It takes someone with expertise to catch those mistakes. And it takes time to correct them.

The question is whether owners of companies will be okay with mistakes being made by AI and less skilled employees, if it saves them a lot of money.

9 replies, 655 views

Thread info Bookmark this thread Trash this thread

Reply to this thread

Always highlight: 10 newest replies | Replies posted after I mark a forum

Replies to this discussion thread

9 replies	Author	Time	Post
Bad news, star employees: You're not the ones who'll benefit the most from AI (Business Insider) (Original post)	highplainsdem	Sep 18	OP
At this point, the errors are yuge security vulnerabilities...	getagrip_already	Sep 18	#1
Isn't this also true with human programmers as well? n/t	Yavin4	Sep 18	#4
Not experienced ones.....	getagrip_already	Sep 18	#6
Eventually, the top employees will be the ones with the ability to understand and train/tune	Yavin4	Sep 18	#2
A company just demonstrated a machine model that does that...	getagrip_already	Sep 18	#7
Most companies are not even remotely close to implementing machine learning tools	Yavin4	Sep 18	#9
AI	NowISeetheLight	Sep 18	#3
Like everything else in life, AI will improve in accuracy.	jimfields33	Sep 18	#5
K&R	Hugin	Sep 18	#8

Response to highplainsdem (Original post)

Mon Sep 18, 2023, 01:18 PM

getagrip_already (11,767 posts)

1. At this point, the errors are yuge security vulnerabilities...

The code generally works, but is neither high performance nor hardened to modern security standards.

Maybe that will improve with time, but using ai code today will inject a lot of vulnerabilities scanning tools may not find, and that script kiddies may not understand.

Reply to this post

Response to getagrip_already (Reply #1)

Mon Sep 18, 2023, 01:22 PM

Yavin4 (34,702 posts)

4. Isn't this also true with human programmers as well? n/t

Reply to this post

Response to Yavin4 (Reply #4)

Mon Sep 18, 2023, 01:47 PM

getagrip_already (11,767 posts)

6. Not experienced ones.....

As a rule, they have been through enough code reviews with security teams to put in the right building blocks.

But once they start relying on ai, and get rid of code reviews, and rely only on qa teams, things will get dicey.

Reply to this post

Response to highplainsdem (Original post)

Mon Sep 18, 2023, 01:21 PM

Yavin4 (34,702 posts)

2. Eventually, the top employees will be the ones with the ability to understand and train/tune

AI deep learning models. You'd be surprised by how much work in corporate America is really rote clerical work that does not require critical thinking. Heck, right now, a good Data team could replace a sizeable portion of the corporate workforce without even using any AI models.

Take lawyers for example. A good portion of their job is drafting briefs/memos based on current case law. A model trained on all existing case law, statutes, prior memos, etc. can easily and quickly draft an early brief. You'd still need a lawyer to look it over and make any corrections. These corrections will be used to update the model.

If I was a young person today, I would study Machine Learning.

Reply to this post

Response to Yavin4 (Reply #2)

Mon Sep 18, 2023, 01:49 PM

getagrip_already (11,767 posts)

7. A company just demonstrated a machine model that does that...

And does it better than you can.

Think you have a future as a prompt designer?

Too late....

Reply to this post

Response to getagrip_already (Reply #7)

Mon Sep 18, 2023, 02:12 PM

Yavin4 (34,702 posts)

9. Most companies are not even remotely close to implementing machine learning tools

Sure. There are demos galore, but day to day implementation is still years off. Heck, most companies aren't even in the cloud for goodness sake.

Reply to this post

Response to highplainsdem (Original post)

Mon Sep 18, 2023, 01:21 PM

NowISeetheLight (3,390 posts)

3. AI

If you're already one if the best some computer program won't help you. Since corporate pay structures seldom recognize differences in performance, these lower skilled but "helped" workers are probably being paid close to what the top performers.

Reply to this post

Response to highplainsdem (Original post)

Mon Sep 18, 2023, 01:23 PM

jimfields33 (14,257 posts)

5. Like everything else in life, AI will improve in accuracy.

It’s still fairly new. We just need to be careful on what jobs to eliminate as AI expands. Obviously, our world has advanced to the point where a person doesn’t have to pour chocolate into a mold for candy or a lamplighter or a switchboard operator or a bowling pin resettler. We’ve always progressed and we’ll never stop doing so.

Reply to this post

Response to highplainsdem (Original post)

Mon Sep 18, 2023, 02:02 PM

Hugin (31,684 posts)

8. K&R

Reply to this post

General Discussion (Forum)