Interesting. However what we are seeing is large scale AI adoption in not only large corporate companies but even medium and small scale industries where it's so much better since labor cost savings are exponential.
If this is for example Google or Palantir, even if they layoff hundreds of thousands of workers in the future, they will retain the best talents to handhold AI while we are in the HITL. But with just one or a few managing small businesses the risk is pretty high that they would be encouraged to move to HOTL directly through a false believe that AI can manage their small level of operations easily.
Sure! I think my thoughts for many decisions should be how it affects the customer. For example, if your agent can scrape internal wikis and send an email to a coworker, great. If it pulls a 4 year old deprecated wikis as a source of truth, coworker can bring it up.
But when it pulls dated docs and sends a fact to a client or customer, I think a human should review and click send. Reputational damage can be just as bad as a data breach.
To your point, I’m really interested in HOW and WHEN the medium and small companies implement HITL. If you have a smaller staff but big aspirations, less HITL might even be an incentive to keep costs low.
In tech, usually the gains trickle down fast. So any agent efficiency and productivity gains in large corpos is bound to be replicated in small businesses. I am very interested in how the latter will manage the increased need of barricading the agent. Google can let the agent roam around its network and get stuff done on its own because it would have the resources to set up guard rails if information were to ever smuggle out. Small companies won't.
So I think, extreme caution is necessary while introducing them into business.
Good write up. I think such essays should be mandatory for students learning to deal with AI since they question how can they best approach the dangers. Well done.
Really appreciate your encouragement Sanjeet. I do agree, large companies have so much more resources they can spin up experimental teams and test the agents full throttle. If it fails the reorg. If a small business tries that and fails, then out of business.
All of your posts are so informative, explaining the "why," as well as the operational specifics and potential ill effects. I appreciate all of your insight for us laypeople
Hey really appreciate it! That means a lot since my “angle” is trying to keep it correct on a technical side but readable and approachable for the general audience.
Interesting. However what we are seeing is large scale AI adoption in not only large corporate companies but even medium and small scale industries where it's so much better since labor cost savings are exponential.
If this is for example Google or Palantir, even if they layoff hundreds of thousands of workers in the future, they will retain the best talents to handhold AI while we are in the HITL. But with just one or a few managing small businesses the risk is pretty high that they would be encouraged to move to HOTL directly through a false believe that AI can manage their small level of operations easily.
Sure! I think my thoughts for many decisions should be how it affects the customer. For example, if your agent can scrape internal wikis and send an email to a coworker, great. If it pulls a 4 year old deprecated wikis as a source of truth, coworker can bring it up.
But when it pulls dated docs and sends a fact to a client or customer, I think a human should review and click send. Reputational damage can be just as bad as a data breach.
To your point, I’m really interested in HOW and WHEN the medium and small companies implement HITL. If you have a smaller staff but big aspirations, less HITL might even be an incentive to keep costs low.
In tech, usually the gains trickle down fast. So any agent efficiency and productivity gains in large corpos is bound to be replicated in small businesses. I am very interested in how the latter will manage the increased need of barricading the agent. Google can let the agent roam around its network and get stuff done on its own because it would have the resources to set up guard rails if information were to ever smuggle out. Small companies won't.
So I think, extreme caution is necessary while introducing them into business.
Good write up. I think such essays should be mandatory for students learning to deal with AI since they question how can they best approach the dangers. Well done.
Really appreciate your encouragement Sanjeet. I do agree, large companies have so much more resources they can spin up experimental teams and test the agents full throttle. If it fails the reorg. If a small business tries that and fails, then out of business.
Extreme caution is the take away.
All of your posts are so informative, explaining the "why," as well as the operational specifics and potential ill effects. I appreciate all of your insight for us laypeople
Hey really appreciate it! That means a lot since my “angle” is trying to keep it correct on a technical side but readable and approachable for the general audience.
You found the mix, and I really appreciate that you make it approachable