Genetic AI mutants; RAGs to riches; CNI Red Teaming and more

Bombs away: It’s been a geopolitically fraught week - again - and so perhaps, like many others, our eyes were a little attuned to that space.

Whether that was an audit of the US Department of Defense highlighting (on page 285 since you ask) how poor the Pentagon’s fragmented and expensive ERP* systems are; the UK’s Ministry of Defence forking out for a synthetic wargame platform; or CISA Red Teaming critical national infrastructure rather badly (well!), security has been on the mind.

*Why are ERP migrations often so damn painful?

Sometimes, Reddit does have the right answer:

I’d say 80-90% go over budget or over time, not do to anything specific to [vendor], but due to the absolute clusterf*ck of people cowboying systems to “just make it work” that its nearly impossible to accurately project. The time it takes to actual unravel wtf each person and department is doing on the user side along with what random ‘fixes’ developers put in that are structurally unstable stresses me out just thinking about it. The only time i’ve seen a project go well is when the client had an A+ db admin that basically ruled with an iron fist.

The report from CISA meanwhile is particularly informative and also somewhat troubling – given that the CNI organisation’s leadership, quote:

Actively “deprioritized the treatment of a vulnerability their own cybersecurity team identified, and in their risk-based decision-making, miscalculated the potential impact and likelihood of its exploitation”.

Much work to do here...

Talking of national security, the late Henry Kissinger and Google founder Eric Schmidt may seem like unusual bedfellows but they wrote a book together... and in it, they have unusual warnings, not least against the “society that chooses to create a hereditary genetic line of people specifically designed to work better with forthcoming AI tools”.?

Don’t say we weren’t warned!

More prosaically, in a posthumously published piece (on Kissinger’s side!) they suggested that “AI could in fact remove humans as a proxy in warfare entirely.” That seems astonishingly unlikely, but form your own views…?

In less fraught territory, it’s earnings season again. Snowflake and Palo Alto Networks have both done well; NVIDIA continues to hoover up billions...

We also like to dip into the non-technology ones to see how more “traditional” enterprises are thinking about IT and innovation.?

Retailers Walmart and Target both reported this week.?

Jasper Hamill looked at Walmart’s approach.

One of the companies heavily involved in retailers innovate with AI is our partner MongoDB. We sat down with Genevieve Broadhead to hear more.

What else was new?

The Bank of England has finally published the results of its AI in financial services survey; our friends at API gateway pioneer Kong have raised a ZIRP-throwback large investment round to expand even faster; the EU’s "GDPR-level" Cyber Resilience Act has come into force; AWS Aurora Serverless is now actually, serverless (history here); CISA boss Jen Easterly is off; Europe’s largest retailer-turned-private-cloud-provider has teamed up with Google to offer a client-side encrypted, “sovereign” Workspace…

And a whole lot more.

Forskolin, luciferase…?

What else have we been reading??

Well, meatspace memory as well as computer memory is always an intriguing area; various scientists have fallen down wayward rabbit holes trying to understand how it works (paging Rupert Sheldrake).?

A new paper in Nature by neuroscientist Nikolay Kukushkin and colleagues at New York University challenges the presumption that memory is stored in the brain. Their study shows that “all cells—even kidney cells—can count, detect patterns, store memories, and do so similarly to brain cells.”

Fascinating.?

That’s it from us this week.

?As ever, we love to hear from readers, so feel free to get in touch on [email protected]. Lots more good stuff in The Stack.

Don't miss out: Get yourself subscribed directly on our website too.