Good article. A little heavy on the hyperbole but mostly tracks. I sat in on a defense conference this week, and back to back PhD speakers talked about AI. Many of the same facts were presented but they disagreed on what level of autonomy from humans would ultimately be achieved. Right now AI is very useful if you can feed it good data and provide clear directions. Will we reach the point where that human direction will no longer be required, something closer to cognitive thought? I don’t think anyone is quite sure where that ends up.

We have a paid AI tool that is quite good for writing, especially things like capability statements and requirements compliance. It still requires some direction because it doesn’t fully understand the source data and cannot differentiate between data from a past proposal and actual program past performance. For us it’s a very useful tool but still requires human oversight.

On the upside, humanoid robot tech is lagging. So it’s not T1000 time just yet.

Lifted from an article on X that Paul posted:

Access to AI is everywhere, and it's free. Start using it. Make yourself spend a few minutes each day learning about AI by using AI.

As far as attempting to reign in AI in any manner, that ship has sailed...

Agreed, this is a key point. It’s a useful tool, those of us that don’t learn to use it will become obsolete.

I spent a good subset of my career in IT automating bigly datacenters, networks, etc. as a systems guy .... GIGO.

And those shades of grey ... between the ears ... it will still matter even with AI ... which is a tool.

Does this phone make me smarter ;)?

I doubt it ... but ... :D

Long ago, early 1990s time-frame I had several computes in my office and it various programs to build computer graphics and output them to photographic film or a large format inkjet printer. Lots of steps to go through, and it was easy for me, as I pretty much built it to work. I remember sitting there thinking someday AI like HAL in 2001 could do all the work as I sit there and just watch. Then I realized that day would be my last as the boss would not need me.

Computers are leaps and bounds more sophisticated and capable now and AI is here for the end user. I see Photoshop 2026 now has some AI tools that I have yet to mess with. I am glad that I am now the boss, and I can't see firing myself. We will need an airplane and humans to do the tasks needed to make our aerial surveys and orthophotos for our surveyor companies.

It's about to become ad supported. Can't wait...

So many potential downsides to the now unstoppable AI which is escalating at a scary pace. The obvious, which the article focuses on, is AI replacing human jobs from blue collar to white but what will that do to economies? No mention of this in the article but most economies of the world (including ours) are dependent on income tax revenue to keep money cycling.

http://forums.pelicanparts.com/uploa...1770924734.gif

Two teams, OpenAI and Anthropic, are in a race to build an AI that transcends the current model for AI and becomes self aware with the ability to understand, learn, and apply knowledge across a wide range of tasks, similar to human intelligence. And this build out is happening with no enforceable safe guards - none.

Just watching the industry, I believe this level of AI, known as AGI, has already been achieved. Is this a good or bad thing? Who knows? time will tell...

In working with AI tools and talking to people way smarter than I am, we aren’t there yet. There are lots of assumptions that the current pace of advancement will continue, which isn’t guaranteed. So we’ll see.

I'll just wait for the nukes to fly and the EMP to end it all.
Until then AI seems to be good for making chicks with 3 boobs
and other important stuff. :D

http://forums.pelicanparts.com/uploa...1770926527.jpg

I lived through the total destruction of the publishing industry as we knew it 40 years ago. People are good at adapting when they have to. It won’t be easy.

I have a well, solar power, 5 acres, and I know how to grow food. I’m reasonably prepared for the AI apocalypse.

The economy regularly goes through massive changes. When the country was using horses to move people and goods around, wheelwrights, barn builders, saddle makers, buggy whips and manure shovels were all booming. Then the car came along and destroyed that industry. in 2025 there were mores horses in the USA than in 1925. Most all used for purely pleasurable purposes.

When telephones were adopted, the telegraph essentially died.

Newspapers are going away rapidly, and the survivors are all brought under one ownership company. Digital papers are pitiful in size compared to the "good ol days" when I was a newspaper photographer.

Robots and CNC totally changed machining.

Entire industries are turned on their heads regularly. I worked at a professional photo lab with 5,000 square feet of floor space in a custom built building with 2 color film processors, and three paper processors, 12 - 4x5 enlargers and two 8x10 enlargers, with a dozen employees. It all just went away due to digital photography. Now in my office I can produce a better tougher, longer lasting final product from my little office, and a short trip to the place that makes my inkjet prints, and just me, not a dozen employees.

AI will change things, but smart humans will figure out legal way to make money in an changing world.

If your turbo is CIS then you can still go for a drive post-EMP!

In the stock market, the "AI replacement" trade is spreading to trucking brokers, freight forwarders, logistics companies, commercial real estate brokers and management companies, insurance brokers, payroll processors . . . where-ever someone can launch a vibe coded "AI agent" that supposedly does the task.

Yup, like startups with an AI agent are going to be hired by commercial property owners to manage or sell their properties, by corporations to handle payroll and taxes, by major shippers to get containers from Vietnam to Frankfurt or trailers from Florida to Michigan, to handle their business insurance and reinsurance layers . . .

AI is most penetrated in coding. How many companies send unreviewed critical code straight from LLM to production?

Most people or businesses conducting high value or non-standard transactions are going to want to deal with a person, face to face. At that point it’s more about trust and human relationships than your ability to process a form quickly. It’s also how we are wired as humans.

That's optimistic. In the pure business models, the end result is achieved at x cost saving over a human then they get a contract. OTOH, we have seen regulation written to force higher prices paid for social justice reasons.

What's worse, society has slowly been conditioned to achieve a quick end result through the apps; fast food/grocery delivery, dating/love, and all aspects of consumerism.

I think the lack of trust of technology paired with scammers expansive and historical use of technology will be the friction point... We have seen how this technology is used nefariously by bad actors and others. Sure, we can choose not to participate but that will be only temporary until the rope is proverbially tightened around our neck.

-AZ

AI is already almost good enough to 100% duplicate your voice with only 10 seconds of you talking. Lots of people have been scammed by AI calls from family members asking for bail money, urgent cash, etc. Ironically that’s going to push some business deals back to good old face to face. Not to mention that AI can’t sell a product that doesn’t exist, such as custom or highly engineered items.

I agree, theory only goes so far...

Too bad we outsourced our industrial base (including skilled laborers) to cheaper regions of the world.

The funny part of this - for me at least - is that I’ve been working on some code that listens to speech, transcribes it and then calls an agent {to do useful things} when it detects commands in the stream.

For the POC I was using “HAL” as a trigger to denote what follows is a command. Yes I’m old. And geeky.

Sadly speech to text kept transcribing it as “How” - and it got so frustrating I switched to “Agent”… it’s quite surprising how easy it is to get this all working when you are used to writing your own io drivers, but even with all this abstraction and sophistication, speech to text isn’t 100%...