I realized senior management didn’t understand our work when they suggested that we use AI to do our stock assessment models and we’re excited because it could automatically fit models to EXCEL spreadsheets. It was awkward when I asked the Microsoft rep during training whether the AI would be verifying if model assumptions were being met. As far as I can tell, the AI just plops on a simple linear regression or LOESS.
The simple fix is, as always, just review the email before sending. No oversight is always a mistake.
Right now i have lengthy conversations, an hour or so. Then i have to summarize the conversation, making note of certain things for compliance. Then next call, i review the summary to get back up to speed.
At some point im going to have a transcript generated, then an ai summarize the call and note compliance points. No harm in that and it will save me an hour some days.
I would probably be less inclined to blame these companies if they didn’t use the very (probably intentionally) confusing phrase “artificial intelligence” and instead called it a “predictive model” or similar. Also if they talk about being on the verge of artificial super intelligence.
1 Like
I have to wonder how much of the market cap is dependent on this.
I need to get a new laptop for work. I provided Microsoft copilot with the website listing my 4 options and asked it to rank them for me and tell me why they were ranked that way. The AI changed the model of one of them for unknown reasons.
I’m not sure that senior managers and CEOs would understand the difference.
1 Like
“You didn’t ask me for your opinion. You asked me for my opinion, Dave.”
2 Likes
Made harder by the fact that we have to pretend that CEOs and other executives are also brilliant, able to understand the deeply technical details of their company’s products. The alternative is that their astronomically high pay isn’t merited, risking the entire story that we live in a meritocracy. So statements by Elon Musk or Mark Zuckerberg on AI become statements by “experts.”
I’ve started working on our AI productivity toolkit for small-mid sized insurers.
Here’s the overview. The push is that the use of AI inhouse right now is non-compliant. Employees are sending internal information to external entities without oversight every time they do a query. The same with custom chatbots, the seed information needs to remain inhouse.
Solution, a selection of various AI’s, hosted in a secure location, access only through an inhouse app, and connects to the various AI’s via a VPN. Basically an internal solution. That lets ee’s use AI for productivity to their heart’s content, no more concerns about internal info being sent to an external entity.
The app will allow user to choose their AI, with different AI’s targetting various needs. I.e. one for summarizing, one for code, etc. And we’ll have one ‘rag’ system where the AI has access to internal documentation to answer things like ‘customer is looking for this form, where to I find it’.
Today I’m working on the reporting functions for the system. After that, I have to determine how to integrate the various AI"s into external (to the AI) apps.
Thoughts on this?
I frequently encounter people who seem to argue that brown people and women are unable to get jobs on merit. Every accident or mishap involving a brown person or woman in charge is due to DEI. Yet they ignore accidents/mishaps with white males in charge.
1 Like
Many of the AI leaders are brilliant experts. Hassabis, LeCunn, Amodei, Sutskever, all are leading scientists with decades of bleeding edge research. And they all talk to the press sometimes, sharing their (sometimes outlandish) opinions.
However, we overwhelmingly prefer to hear from a lying billionaire with a nice smile than some nerd with weird hair and a stutter. So, Musk, Zuckerberg, and Sam Altman get more power, money, and press.
AI fails yet again…
I gotta haul ass and get some work done - busy day today and instead of working I’m perusing the web for hardware to set up a home AI server lol.
sounds like a different kind of risk posed by ai.
