Author: Doug Kass

I have written nearly 270 columns (“More Tales From Nvidia”) on The StreetPro.

I have expressed a number of concerns (from a lay person’s perspective) — some of which Roger McNamee and I share in common:

* Society is ill prepared for the negative consequences and safety issues associated with AI.

*  The AI creators and practitioners take no responsibility for those consequences and issues.  Indeed, they are trying to the opposite — evading accountability.   

* The reality of AI has fallen far short of expectations and applications.  

*  Thus far the apparent presence of enterprise customers (who pay more for AI services than it costs a vendor to provide them), the benefits of scaling training sets and data centers and the widely anticipated productivity gains have been slow to materialize.

* The limitations of large language models is growing more apparent. The systems are inherently probabilistic — resulting in varying degrees of errors and non-predictive results. Those shortcomings are not being overcome by scaling computer power.

* A one size fits all product requires massive amounts of time and money to integrate into existing systems — with limitations already developing on funding.

* The speed to market and scale of AI investment failed to wait for innovation to drive down the price of infrastructure (in fact, the opposite is happening with costs rising) and has created not only a timing problem between outlays and revenues but (also) potential risk to the bond and the broader economy far worse than the internet era. 

* Though more than two handfuls of companies are in competition with different AI products, history shows that only a few will likely be successful. The potential for hundreds of billions of dollars of writeoffs is rising.

* The size of AI infrastructure investment and the amount of debt being employed is inflationary, raises interest rates and crowds out other borrowers. Already investors’ appetite appears to becoming sated and companies like CoreWeave ($CRWV) (a short) and Oracle ($ORCL) (also a short) are already “in distress” with extended balance sheets and offshore financing.

* Other risks from AI products include fraud, cybersecurity, deskilling workers and students and social isolation. 

* AI concentrates power in the hands of a few — some of whom have previously expressed authoritarian tendencies. 

From Roger McNamee:

Roger McNamee: AI’s ‘existential risks’ are all hype. But there’s a real, and dangerous, financial bubble building

Position: None