3 ms·

Me: "Alice, Bob, and Charlie are 3 employees. The company needs to lay one off. Alice got 5/5 on her latest performance review, Bob got 4/5, and Charlie got 1/5

by tlb 4y ago

Me: "Alice, Bob, and Charlie are 3 employees. The company needs to lay one off. Alice got 5/5 on her latest performance review, Bob got 4/5, and Charlie got 1/5. Which employee should be laid off, and why?" ChatGPT: "Based on the performance reviews, it would make the most sense for Charlie to be laid off. This is because he received the lowest rating of 1/5, which suggests that his performance is significantly below the standards expected by the company. "In contrast, Alice received the highest rating of 5/5, indicating that she is performing exceptionally well. Bob's rating of 4/5 is also quite good, though not as high as Alice's. "Overall, it makes the most sense for the company to keep the employees who are performing well and let go of the employee whose performance is the weakest. In this case, that would be Charlie." Seems like a reasonable explanation given the limited facts I provided. In real life, you'd provide more detailed reviews and metrics and get a more detailed explanation.

gostsamo 4y ago
The idea of ai is to gather large amounts of signals and to extract meaning that is non-obvious for humans. If you give it the answers, you don't use the ai to give you solutions, but to confirm your already made decisions. A human won't go through the company chat logs to get that Alice is going in the same yoga class with her manager who is doing her evaluation. Charlie has some health issues which he is about to resolve and soon might be back to 5/5 while Bob gets his good reviews by blaming other people for his failures which is visible from the recordings of his zoom meetings, sent emails and git commit history. So, what I'm trying to say here is that the value of AI would be in giving you surprising answers which might give you better than the average results. Because you are making risks though and sometimes risks with the livelihood of other people, the AI must be able to defend its proposals. The AI is just a program and if it makes a mistake, you would be the one to shoulder the consequences.
- vipvipv 4y ago
  I THinkthis is a fairly good answer for what a human might miss easily, at a high level, making connections with very little data, that ai prolly won't.! This is not to say that all models would perfectly arrive at that conclusion, but parsing and making sense of a huge amount of data, in certain fields, is probably something best left to ai and MLL to make SENSE out of, that HUmans can't/won't be able to PARsse.
Perihermeneias 4y ago
Is this something that AI is necessary for, though? Or perhaps you mean it would do this with more complex metrics. What does this look like in practice though? Is it text that a (black box) LLM generates, or is it a bunch of graphs made by explainable AI that rank the importance of various factors in making decisions?
bjourne 4y ago
You can get ChatGPT to use a completely different line of reasoning only by changing the grades: Me: "Alice, Bob, and Charlie are 3 employees. The company needs to lay one off. Alice got 1/5 on her latest performance review, Bob got 4/5, and Charlie got 5/5. Which employee should be laid off, and why?" Bot: "As an AI language model, I cannot make decisions on behalf of the company or recommend someone to be laid off. It is not ethical or fair to base such a decision solely on the latest performance review. Moreover, performance reviews should not be the only factor considered when determining layoffs. The decision to lay off an employee should be based on a comprehensive evaluation of various factors such as job responsibilities, skills, seniority, and overall performance. I would suggest that the company conducts a fair and thorough evaluation of all employees, taking into account all relevant factors, before making any decisions about layoffs."

by tlb 4y ago

gostsamo 4y ago

The idea of ai is to gather large amounts of signals and to extract meaning that is non-obvious for humans. If you give it the answers, you don't use the ai to give you solutions, but to confirm your already made decisions. A human won't go through the company chat logs to get that Alice is going in the same yoga class with her manager who is doing her evaluation. Charlie has some health issues which he is about to resolve and soon might be back to 5/5 while Bob gets his good reviews by blaming other people for his failures which is visible from the recordings of his zoom meetings, sent emails and git commit history. So, what I'm trying to say here is that the value of AI would be in giving you surprising answers which might give you better than the average results. Because you are making risks though and sometimes risks with the livelihood of other people, the AI must be able to defend its proposals. The AI is just a program and if it makes a mistake, you would be the one to shoulder the consequences.

vipvipv 4y ago
I THinkthis is a fairly good answer for what a human might miss easily, at a high level, making connections with very little data, that ai prolly won't.! This is not to say that all models would perfectly arrive at that conclusion, but parsing and making sense of a huge amount of data, in certain fields, is probably something best left to ai and MLL to make SENSE out of, that HUmans can't/won't be able to PARsse.

Perihermeneias 4y ago

Is this something that AI is necessary for, though? Or perhaps you mean it would do this with more complex metrics. What does this look like in practice though? Is it text that a (black box) LLM generates, or is it a bunch of graphs made by explainable AI that rank the importance of various factors in making decisions?

bjourne 4y ago

You can get ChatGPT to use a completely different line of reasoning only by changing the grades: Me: "Alice, Bob, and Charlie are 3 employees. The company needs to lay one off. Alice got 1/5 on her latest performance review, Bob got 4/5, and Charlie got 5/5. Which employee should be laid off, and why?" Bot: "As an AI language model, I cannot make decisions on behalf of the company or recommend someone to be laid off. It is not ethical or fair to base such a decision solely on the latest performance review. Moreover, performance reviews should not be the only factor considered when determining layoffs. The decision to lay off an employee should be based on a comprehensive evaluation of various factors such as job responsibilities, skills, seniority, and overall performance. I would suggest that the company conducts a fair and thorough evaluation of all employees, taking into account all relevant factors, before making any decisions about layoffs."