Learning Library

← Back to Library

Human Skills That Outlast AI

Key Points

  • Skills centered on nuanced, in‑person human interaction—such as empathy, care, and real‑time feedback—will remain valuable despite AI advancements.
  • AI can generate recipes but lacks the sensory feedback loop that human chefs use, leading to less nuanced and less preferred dishes in blind taste tests.
  • While AI is rapidly improving at medical diagnosis and may soon surpass doctors in accuracy, the human presence of nurses and clinicians will become crucial for delivering compassionate bedside care.
  • The future job market will favor roles that emphasize human qualities (e.g., kindness, bedside manner, relationship building) that AI cannot replicate, a theme explored in the speaker’s “AI and Careers” course.

Full Transcript

# Human Skills That Outlast AI **Source:** [https://www.youtube.com/watch?v=u9nlWdWFoVw](https://www.youtube.com/watch?v=u9nlWdWFoVw) **Duration:** 00:06:52 ## Summary - Skills centered on nuanced, in‑person human interaction—such as empathy, care, and real‑time feedback—will remain valuable despite AI advancements. - AI can generate recipes but lacks the sensory feedback loop that human chefs use, leading to less nuanced and less preferred dishes in blind taste tests. - While AI is rapidly improving at medical diagnosis and may soon surpass doctors in accuracy, the human presence of nurses and clinicians will become crucial for delivering compassionate bedside care. - The future job market will favor roles that emphasize human qualities (e.g., kindness, bedside manner, relationship building) that AI cannot replicate, a theme explored in the speaker’s “AI and Careers” course. ## Sections - [00:00:00](https://www.youtube.com/watch?v=u9nlWdWFoVw&t=0s) **Human Skills That Outlast AI** - The speaker argues that interpersonal, sensory, and feedback‑driven abilities—exemplified by cooking recipes and nuanced medical diagnosis—will remain valuable because AI lacks genuine human experience and nuanced feedback loops. ## Full Transcript
0:00if you're wondering what skills are 0:01going to persist in a world of AI I 0:04would challenge you to think about what 0:06skills are most valuable for human 0:08in-person 0:09interaction those kinds of skills are 0:12not going to go away and yes they can be 0:15transferable to remote work environments 0:17Etc but the thing to call out is there 0:20are certain modalities of Being Human 0:23that AI is not very good at I'll give 0:25you an 0:26example recipes 0:30AI can make 0:31recipes but pretty consistently so far 0:35in blind taste tests humans prefer human 0:38generated 0:39recipes AI tends to overemphasize a 0:42particular flavor profile AI tends to be 0:44less 0:45nuanced and I believe this is because AI 0:48doesn't have an effective feedback loop 0:50AI doesn't taste AI doesn't bake 0:53something and say m this cookie isn't 0:55quite right it's just a little bit dry 0:57let me see if I can modify it humans do 1:01and the best human chefs are going to be 1:04repeating and getting that feedback loop 1:06going with their own taste Buzz with the 1:08taste buds of their customers until they 1:09get something exactly right that doesn't 1:12mean we're all going to become chefs in 1:13the age of AI but I think it's a good 1:15example I'll give you another example AI 1:18is better and better and better at 1:22medical 1:23diagnosis every single model so far 1:26released in 2024 has been incrementally 1:29better 01 was a big jump in medical 1:32diagnosis capabilities that doesn't mean 1:34that we're using it for medical 1:36diagnosis but it does mean that it is 1:38getting better and better at answering 1:40medical diagnosis type questions 1:44accurately and with that in 1:46mind I think that we are going to start 1:48to see a shift over the next 10 15 years 1:51toward registered nurses and nursing 1:54type professions as being very very 1:56valuable for humans because the human 1:58presence matters a lot 2:00in providing medical care but the actual 2:03diagnosis we may be getting to a point 2:05where the AI is actually better at 2:08diagnosing humans with diseases than 2:11doctors are and so again it's a 2:13situation where you might 2:15think that the doctor is the one that 2:18retains the power in the scenario 2:20because doctors have more knowledge 2:22doctors have to go to school for more 2:23years it's not necessarily true I think 2:26we're getting to a point where the value 2:28of incremental knowledge is not 2:29necessarily is obvious because the AI 2:31can learn so quickly could the AI do a 2:34doctor's job today from a diagnosis 2:36perspective not quite not really in five 2:40years I wonder I wonder and if that's 2:43the case then that's going to put value 2:45back on bedside manner it's going to put 2:47value back on humans being kind and 2:49caring and the health care experience 2:51things that the AI just is not going to 2:53be able to do as effectively the robots 2:55that come and care for you are not going 2:56to provide the level of human care and 2:58concern that a person can 3:02so when I think about that I also think 3:04about the professional world where can 3:07we look in our professional lives for 3:09human skills that are going to stay with 3:12us this is part of what I'm covering in 3:14the uh Ai and careers course that I'm 3:16doing on 3:17Maven if you look at your career as a 3:22series of investments in long-term 3:25payouts from a skill perspective I 3:27believe that there are strong long-term 3:29payouts still ahead for human skills and 3:32I think the technical skills are leaning 3:34more on AI intelligence allocation and 3:37they're leaning more on how you think 3:39about applying artificial intelligence 3:42to business problems that have a human 3:44component for instance the AI might have 3:47thought that Google's ads describing a 3:50child using a large language model to 3:52write to their favorite track star in 3:54the Olympics would be a good idea it's 3:57topical it talks about the capabilities 3:59of a large language model correctly and 4:01accurately it's got a cute kid 4:04angle but as a human it really fell flat 4:08it fell flat because we associate the 4:10idea of writing a letter to a star from 4:13a kid as a uniquely human activity where 4:16we can see the kid using the pencil to 4:18form the letters and they're doing their 4:20very 4:21best and it just didn't work and 4:23eventually Google had to pull the 4:25ad in those situations the human 4:28perspective on what humans like and what 4:31humans want is going to remain 4:34critical I'll also go into the technical 4:36side of things I think that there will 4:38remain a ton of value in humans 4:41understanding and architecting large 4:43software systems over time so if you're 4:45in the developer 4:47space perhaps the value of pumping out 4:50code that is often written before and 4:52therefore in training models is going 4:54down but the value of solving novel 4:56problems as an engineer in an efficient 4:59way 5:00is something I think that will take 5:02longer for the AI to effectively address 5:05and I suspect that that is because 5:07humans have particular behavioral needs 5:11that other humans understand at scale 5:14and can best represent with engineering 5:16practices so if you're trying to 5:18understand how 5:20cdns actually respond during a prime 5:23time football streaming event which is 5:26something we had to deal with at Prime 5:28video that is going to depend on a 5:30knowledge that people are more likely to 5:34tune in and start streaming the game if 5:36it is close if it is in the fourth 5:38quarter if it is going to overtime 5:40things like that whereas from an llm 5:43perspective from an AI perspective even 5:45if it's not an llm all you're going to 5:47see is a series of previous traffic 5:50Events maybe that's enough maybe that's 5:54enough to design the system but I would 5:56challenge you and I would say You're 5:59going to to get better systems long term 6:02with that human perspective with an 6:04understanding of why the software is 6:06being used the way it is so there you go 6:09just a few little thoughts that I had on 6:11how humans can build long-term durable 6:13skills and areas where humans are 6:15continuing to Excel and I think are 6:17likely to continue to excel in the age 6:19of 6:20AI where we have feedback loops that are 6:23uniquely human we are going to remain 6:26strong contenders for highly skilled 6:29workers professions places where we can 6:32add value to the 6:33market and that doesn't mean by the way 6:36that I don't see a lot of potential for 6:37AI of course I do it just means that we 6:40need to learn where is AI going to be 6:42good and where is AI going to be good by 6:45working with us so on to the Brave New 6:48World Together