I surely hope you’re right, @Data10, but the labor economy is only one area potentially threatened by AI. Our national security is another. What bad actors can do with this technology on the military front may make our (any) economy moot.
1 Like
As I’ve mentioned, ShawSon who is a denizen of techland has believed that AI would conquer all for 10 years and has been very pessimistic that our polity will be able to make the kinds of changes that will be needed. I’ll have to ask my DIL, who has a senior position at a big tech company that probably is well-situated for an AI-dominant world, how she is thinking about the future.
I just had a meeting with a firm we are contracting with that in a month used network analysis and AI to figure out what has been probably taken us a year to understand. Thus far nothing new, but very impressive. Tomorrow we will dig deeper and they are going to give us a very focused AI agent that they built for the analysis. I’m learning so much.
2 Likes
Anthropic hasn’t released their latest version yet because it will make cybercrime too easy.
2 Likes
I also work in tech. The risk I’m personally most concerned about is a stock market bubble. I see many similarities with the 90s Dot Com bubble, with a decade of returns far above historical norms, driven by high future expectations and high future valuations of the companies supporting the new technology in relation to current earnings. If those high expectations seem unlikely to come to pass and market decides that AI is overhyped, there is a potential for a correction like occurred in 2000s with Dot Com crash, with severe losses that take many years to recover from.
4 Likes
I don’t think I agree with this. Policy moves rapidly– too rapidly in some cases, OR at a snail’s pace. The states expanding the “religious exemption” for non-vaccinators? Moving rapidly with absolutely no thought or research or analysis on who will be impacted, what the societal costs will be of annual measles epidemics, how quickly the immuno-compromised will be adversely impacted. The movement for “clean food” which somehow has incorporated a movement to remove fluoride from drinking water (a massive public health triumph), go back to unpasteurized milk (the dumbest idea in a long time) and increase the recommendations on meat, cheese and egg consumption (another dumb idea)?
I think it’s hard to predict which policy issues will move like wildfire due to shrewd and in some cases self-dealing social media figures, and which ones will sit around in the file cabinet of a think tank and never see the light of day.
But plenty of policy decisions are made with lightening speed and zero consideration given to the societal consequences.
4 Likes
Social media has rapidly expanded the spread of disinformation about issues like vaccines and fluoride. New moms are looking to social media “influencers” and various quacks instead of listening to their pediatrician. It’s the slow death of expertise. So, yes, you are right that some policies do change quickly (but why does it always have to be the bad things).
3 Likes
A bit off topic because it hasn’t necessarily taken their tasks but 20-30k layoffs at Oracle so they can build out their network is truly stunning.
1 Like
Because people trying to make good change are more likely to deliberately study the proposed changes and effects, and at least find out who may be or believes will be affected, while those making bad changes are more likely to make the changes without careful consideration.
Also, in general, it is harder to build something than to destroy it, whether the something is an idea like general trust or a physical object like a building.
2 Likes
Cybercrime is already too incredibly easy. 
(Our son is at the table with Anthropic over the use of their LLM for military purposes. Anthropic wants to withhold…what is already being used.)
3 Likes
I appreciate you sharing these articles. After reading the following paragraph, which comes early in the Open AI piece, tears were welling up in my eyes.
The Case for a New Industrial Policy. Society has navigated major technological transitions before, but not without real disruption and dislocation along the way. While those transitions ultimately created more prosperity, they required proactive political choices to ensure that growth translated into broader opportunity and greater security. For example, following the transition to the Industrial Age, the Progressive Era and the New Deal helped modernize the social contract for a world reshaped by electricity, the combustion engine, and mass production. They did so by building new public institutions, protections, and expectations about what a fair economy should provide, including labor protections, safety standards, social safety nets, and expanded access to education.
I was unable to continue reading much past this, because thinking of how past generations overcame technological disruptions and how our current political system works, I don’t see that much hope.
If people do know of actionable steps that people can take (apart from reaching out to our Congressional repesentatives) to advocate and address these issues, please share. I’m already convinced of the significant changes that AI will bring and of the potential (and even likelihood) of significant harm that will impact much of society. At the moment, though, I alternate between wringing my hands and sticking my head in the sand.
3 Likes
Sorry. Not cybercrime (although that will get easier too) but cybersecurity.
1 Like
It sounds like Oracle is an unique position. Oracle’s stock has lost more than half its value over the past 6 months, and they are over $100 billion in debt – likely a larger debt than any other AI company, with interest rates premiums on the debt doubling over past 6 months. However, Oracle’s saving grace is an extraordinary $550 billion waiting in unfulfilled future contracts for AI infrastructure. That half-trillion in future AI contracts seems like the only viable path going forward, so it seems reasonable to focus the company on that goal. Laying off employees not required for the AI infrastructure contracts helps generate the cash flow needed to continue, as more debt is likely not practical. It’s a big bet that could go very well or very poorly for the company.
2 Likes
From that article:
Leaps in AI model capabilities have come with concerns about hackers using such tools for figuring out passwords or cracking encryption meant to keep data safe.
I posted upthread about the danger of Q-Day (Quantum Day), the uncertain future date when quantum computers become powerful enough to break current encryption standards rendering most modern cybersecurity vulnerable, effectively creating a universal pick lock capable of taking down our entire system of security for banking, bitcoin wallets, electrical grids, privacy data, top-secret military information, and all the rest. We don’t need Anthropic’s Mythos to find this “crack” — it’s a major threat that is widely known. Q-Day is a perilous AI-adjacent threat that is consuming national security resources. I worry more about the ramifications of this issue than what AI might potentially do to our labor economy.
4 Likes
That’s why the stock is down - I own it. But in essence, these folks and it’s a lot of them if the analyst is to be believed, have been harmed by AI. If it’s Oracle, who is next?
Regarding who is next, this touches on my earlier concerns about a potential AI bubble. I mentioned that Oracle has over $100 billion in debt and is laying off employees in order to get the revenue to complete their $550 billion in future AI infrastructure contracts.
These future AI contracts include things like OpenAI agreeing to a $300 billion contract with Oracle to create AI infrastructure for their company, such as massive data centers. This raises the question of how a company that is hemorrhaging money like OpenAI, with tens of billions in net losses, can afford a $300 billion future contract with Oracle and $1.5 trillion in total future contracts? The answer is taking on more debt and bet on rapid future growth, expecting that they will become profitable in the future, when they are bigger. OpenAI has ~$100 billion debt , in order to support current operations that are orders of magnitude lower than what is needed/expected.
If this expectation for extremely rapid growth and more debt doesn’t happen, then OpenAI may be unable to fulfill their future $1.5 trillion in contracts, which spirals Oracle further down beyond their current >50% share price loss over past 6 months, as Oracle is shifting the company towards an all-in bet on future revenue from those contracts. OpenAI also has $500 billion in future contracts with Nvidia. Nvidia is not all-in like Oracle, but adjustments to the $500 billion contract would have a big impact. Nvidia is currently the highest market cap company in the world and highest weighted company in S&P 500. If Nvidia, Oracle, OpenAI, and others all take a major hit, it is likely to spook the market, in particular question whether the high future expectations of AI-focused companies are accurate, which makes acquiring debt more challenging and expensive for those companies. This is the point where I’d expect to see large layoffs at a wide variety of AI companies.
This possible future is by no means guaranteed, but I think there is a realistic possibility.
5 Likes
I’d be interested to know more about how quality / testing / integrity (SOD - separation of duties) is handled with AI. I spend the 2nd half of my career in mainframe software development world, very careful project planning and strict business controls. Hoping same holds true with AI software development.
Our son runs a development team that almost exclusively uses AI for coding, but the team runs that output against the same testing and control protocols required in the past. The origin of the software does not negate the need for those integrity measures. It just takes fewer people end-to-end and iterative changes are lightening fast. The code is still carefully reviewed, and the final product still has to adhere to the spec and pass every quality test. Properly managed, AI boosts quality, reduces errors, and speeds testing, shortening the spec-to-delivery cycle significantly.
ETA: I got a very solid response from asking your question of ChatGPT (How are quality, testing, and integrity (SOD - separation of duties) handled in an AI software development shop?) that describes in detail how AI software development works much better than my short response above. If you have access to Chat (or another LLM), it’s worth reviewing the answer but is too granular to post here.
2 Likes
My husband does a lot with paid chatgpt. But… that does add another question on validity of the info compiled.
Honestly most chatgpt answers have been helpful, seem factual. But I had years of vigilance on SOD etc for closed loop systems - example, developers (the programmers with intimate knowledge of the code, any potential loopholes) don’t have write access on production systems. Curious about how to keep misinformation and fraudulent sources being harvest by the efficient/impressive (and sounds like well managed) AI systems.
Given that we can’t keep any bad actor from harvesting whatever they want from any system, it’s not reasonable to expect we will be able to control AI either (says this former Unix systems administrator).
2 Likes
Software integrity is far from my sphere of knowledge but I think I have read that some companies will use other AI to check their code from Claude (or whatever they are using).
1 Like