• Thanks for stopping by. Logging in to a registered account will remove all generic ads. Please reach out with any questions or concerns.

Trump administration 2024-2028

Democrat victories after the midterms have forced Wall Street to adjust to what they believe will be increased regulatory scrutiny and slowdown in data centre construction.

Temporary blip. I'm in buy mode!
I think that’s part of it; it’s going to curtail Trump’s irrational exuberance and fanboying of something he doesn’t understand.

I think that this is also forcing a more across the board assessment of basics like cash flow and profitability projections. Tons of people and companies can build tools with AI, but how many of those tools will profitably support a business model and underpin publicly traded valuation? I can’t go anywhere online without being spanned by slop ads for slop products, but how many niche apps will survive time in the market with payrolls due biweekly?

There’s also the reality that, regardless of regulatory approvals for data centres, supply chains are jammed up by backlogged demand. ASML can only build a finite amount of extreme-UV lithography machines, and they’re the only real player in that space. That puts a hard cap on how many silicon wafers TSMC and other chip fabs can process into the highest quality integrated circuits. That caps how many bleeding edge GPUs can be output, in turn capping capacity for data centre-scale processing rack modules. Someone starting to build a new data ventre now is in line for compute. You can build a steel shell full of empty server racks easily enough. Can you get the Vera Rubin modules to mount though?

And the capital infrastructure to open up any of this is immense. There’s no easy tap to turn to increase flows at the headwaters of the physical tech stack.

So companies like Nvidia are making out like absolute bandits because they produce the end use product everyone wants. But that means those compute costs are a staggering expense for anyone looking to build their physical infrastructure up.

Strategically this also puts governments in the queue for the same hardware when we talk about wanting data sovereignty, particularly our own compute. Governments, however, have greater capacity to eat the costs and fund the purchases, whereas private sector companies have taken on such dizzying corporate debt that they’ve managed to skew bond markets, and actually drive borrowing costs up for government who have to compete for borrowers.

What a wild time.
 
  • Like
Reactions: ytz
I was talking with a friend that works for a company that does online and in person HR training events worldwide. She was telling about all the areas where they have been implementing AI agents to run their processes and how wonderful it is. I asked her if their annual sales have increased. No.

I worry that in many business cases AI may be able to reduce staffing overheads but won't generate more customers and revenue. What does that mean for the economy as a whole where business output capacity remains the same but the unemployment rate rises?
 
I was talking with a friend that works for a company that does online and in person HR training events worldwide. She was telling about all the areas where they have been implementing AI agents to run their processes and how wonderful it is. I asked her if their annual sales have increased. No.

I worry that in many business cases AI may be able to reduce staffing overheads but won't generate more customers and revenue. What does that mean for the economy as a whole where business output capacity remains the same but the unemployment rate rises?
Depends. I think some of the value enhancing use cases will be ones where job tasks that require human knowledge and intervention can be done more efficiently using AI tools to speed things along. We have several people in the site working in roles where they’re described using AI tools to speed up some of their work processes and to speed up outputs. Calculation, manipulation and presentation of data is one really obvious area; my limited dabbling in this with some simple data sets has really impressed me with how quickly I can dump some data in and get an actually useful output that helps me understand it.

In my work with complex investigations where humans presently pore over the contents of a dozen or more electronic devices at once to find criminal evidence, me and one of the other nerds in the office shake our head at how archaically we still do it knowing what the potential for evidence and intelligence analysis is if we can train the models and host them on secure architecture. We won’t need any future investigators, but we might get preliminary results from evidence analysis in hours or days that take weeks or months currently. Another use case I’ve seen is where AI transcription of audio has finally gotten really good; in my team we’re using an approved AI tool for audio transcription where we’ll get a ‘quite good’ version in minutes when before we’d send it out to human transcribers and see it back in weeks. In both cases an investigator has to do a final review and proof it anyway, so now we can do it all in house and the investigator makes the minor necessary corrections right in the interface. We can get this done and disclosed, or in the hands of our warrant writers very quickly.

I bet just about anyone in any field could probably turn their mind to this and find ways where work processes could be greatly improved to get more or faster output, but that will still need humans in the loop. Productivity gains are there to be had.
 
I worry that in many business cases AI may be able to reduce staffing overheads but won't generate more customers and revenue. What does that mean for the economy as a whole where business output capacity remains the same but the unemployment rate rises?
Reducing costs is roughly interchangeable with increasing profits, and probably more useful since a tighter operation is more likely to have a lower threshold of failure during downturns.

What you describe is unambiguously a good thing - productivity improvements free up labour for other uses, which is essentially how new wealth is created. It's immediately painful for those turned out, but people laid off do not remain out of employment indefinitely (otherwise, pretty much everyone today would be unemployed when compared to all the jobs that have permanently gone away over the past century-plus).
 
Another day, another potentially major hack. This time it’s the FBI’s National Crime Information Centre. This is the federal database for stuff like arrest warrants, release conditions, police caution flags and such. Probably lots of other info on there of some sensitivity.


Non-paywall:
 
  • Wow
Reactions: ytz
I bet just about anyone in any field could probably turn their mind to this and find ways where work processes could be greatly improved to get more or faster output, but that will still need humans in the loop. Productivity gains are there to be had.
There will be a lag, I'm sure.

Never underestimate the human ability to ignore new technology that threatens their livelihood. Back when I was running our law firm's computer systems, at the turn of the century, I was trying to fire up interest in document assembly software. Law firms - solicitors and litigators alike - ran on paper and cut and paste processes that reused the same standard document, slightly modified, for most workflows. A properly set up document assembly program could take much time and error out of the process. The key is "properly set up." That requires upfront work to produce a viable system. If done right it would also be a commercially viable system. BUT. No one wanted to put in the up front time. Off-the-shelf products rarely worked out of the box without heavy customization.

It was the same years later as I was working with JAG on a knowledge/case/records management system. Knowledge management needs knowledge to manage. You first need to assemble that relevant knowledge base. As usual, folks were more concerned about investing their time to manage today's alligator problems than spending time on draining the swamp.

If I look around now I see dozens of legal document management solutions touting their AI engines. None of them tell you what is needed in order to turn their system into a workable solution for a given office. - or better yet, the business transformation that a firm needs to commit to implement the solution amongst all their employees.

Progress needs a heavy up-front investment in time and effort (and cash).

:cool:
 
I worry that in many business cases AI may be able to reduce staffing overheads but won't generate more customers and revenue. What does that mean for the economy as a whole where business output capacity remains the same but the unemployment rate rises?

We're seeing this now. Entry level job market has absolutely crashed everywhere. In Canada, it's at a level not seen since the 80s.

It's a question of distributing the gains of productivity that arise from what is effectively the Fourth Industrial Revolution. Right now the gains from AI, Machine Learning, 3D printing, etc primarily go to capital. It's driving insane wealth inequality. And to the extent that it benefits workers, it's benefiting older workers while decimating young people by making it harder and harder for they to get on the ladder.

A lot of the backlash we see on both where left and the right are largely about trying to grapple with this problem.
 
We're seeing this now. Entry level job market has absolutely crashed everywhere. In Canada, it's at a level not seen since the 80s.

It's a question of distributing the gains of productivity that arise from what is effectively the Fourth Industrial Revolution. Right now the gains from AI, Machine Learning, 3D printing, etc primarily go to capital. It's driving insane wealth inequality. And to the extent that it benefits workers, it's benefiting older workers while decimating young people by making it harder and harder for they to get on the ladder.

A lot of the backlash we see on both where left and the right are largely about trying to grapple with this problem.
They better figure it out quick. The USA has surpassed the estimated level of inequality that was present during the lead-up to the French Revolution. Heads will roll if the youth work but cant survive. The decimation of the entry-level position leaves a ton of shitty low-wage jobs and increasingly scarce high paying ones. Not a good combo.
 
Back
Top