Yup, that is the gist of it. And it seems that people are starting to wake up. You see the biggest issue I have had with any mention of AI, is that it doesn’t (yet) exist. People can shout AI on every corner, but soon the realisation comes in that they were wrong all the time will hurt them, it will hurt them badly. And this is merely a sideline to the issue. The issue is Microsoft and lets get through some articles.
1. Microsoft says cyber-attack triggered latest outage
The first one is (at https://www.bbc.com/news/articles/c903e793w74o) where we see “It comes less than two weeks after a major global outage left around 8.5 million computers using Microsoft systems inaccessible, impacting healthcare and travel, after a flawed software update by cybersecurity firm CrowdStrike. While the initial trigger event was a Distributed Denial-of-Service (DDoS) attack… initial investigations suggest that an error in the implementation of our defences amplified the impact of the attack rather than mitigating it,” said an update on the website of the Microsoft Azure cloud computing platform.” The easiest way of explaining this is to compare Azure to a ball. A foot ball has (usually) 12 regular pentagons and 20 regular hexagons. They are stitched together. Now under normal conditions this is fine. However software is not any given shape, implying that a lot more stitches are required. Now consider that Microsoft 365 is used by over a million corporations. Now consider that a lot of them do not use the same configuration. This implies that we have thousand of differently stitched balls and the stitches is where it can go wrong. This is where we see the proverbial “the implementation of our defences amplified the impact of the attack rather than mitigating it” Microsoft has been so driven by using it all, that they merely advance the risk. And it doesn’t end here. CrowdStrike is another example. We see the news and the fake one person claiming responsibility for it. Yet the reality is that there is a lot more wrong than anyone is considering. These two events pretty much prove that Microsoft has policy and procedure flaws. It is easy to blame Microsoft, but the reality is that we see spin and the trust in Microsoft is pretty much gone. People say “Microsoft’s cloud revenue was 39.3% higher”, yes this is the case, and considering that Amazon was originally a ‘bookshop’, so they went against the larger techies like IBM and Microsoft and they got 31% of the global market share. Not bad for a bookshop. And the equation gets worse for Microsoft, these two events could cost them up to 10% market share. In which direction these 10% go is another matter. AWS is not alone here.
I was serious about not letting Microsoft near my IP. I had hoped that Amazon would take it (they have the Amazon Luna) but it seems that Andy Jesse is not hungry for an additional 5 billion annually (in the first stage).
And as Microsoft adds more and more to their arsenal these problems will become more frequent and inflicts damage on more of their customers. Do I have evidence? No, but it wasn’t hard and my example might give you the consideration to ponder where you could/should go next.
2. Microsoft Earnings: Stock Tanks As AI Business Growth Worse Than Expected
In the second story we see (at https://www.forbes.com/sites/dereksaul/2024/07/30/microsoft-earnings-stock-tanks-as-ai-business-growth-worse-than-expected/) that Forbes is giving us “shares of Microsoft cratered about 7% following the earnings announcement, already nursing a more than 8% decline over the last three weeks” with the added “Microsoft’s crucial AI businesses was worse than expected, as its 29% growth in its Azure cloud computing unit fell short of projections of 31%, and sales in its AI-heavy intelligent cloud division was $28.5 billion, below estimates of $28.7 billion” As stated by me (as well as plenty of others) there is no AI. You see AI would give the program thinking skills, they do not have any. They kind of speculate and they have lots of scenario to give you the conditional feeling that they are talking “in your street” but that is not the case. For this simple illustration we get Wired (at https://www.wired.com/story/microsoft-ai-copilot-chatbot-election-conspiracy/) giving us ‘Microsoft’s AI Chatbot Replies to Election Questions With Conspiracies, Fake Scandals, and Lies’, so how does this work? You see the program (LLM) looks at what ‘we’ search for, yet in this the setting is smudged by conspiracy theorists, troll farms and influencers. The first two push the models out of synch. Wired gives us “Research shared exclusively with WIRED shows that Copilot, Microsoft’s AI chatbot, often responds to questions about elections with lies and conspiracy theories.” Now consider that this is pushed onto all the other systems. Then we are treated to “Microsoft’s AI chatbot is responding with out-of-date or incorrect information”, so not only is the data wrong, it is out of date, as I see it what they call ‘training data’ is as I see it incorrect, out of data and unverified. How AI is that? A actual real AI is set on a Quantum computer (IBM has that, although in its infancy) a more robust version of shallow circuits (not sure if we are there yet) and is driven not by binary systems but framed on an Ypsilon particle system, which was proven by a Dutch physicist around 2020 (I forgot the name). This particle has another option. We currently have NULL, Zero and One. The Ypsilon particle has NULL, Zero, One and BOTH. A setting that changes everything.
But the implementation into servers is to be expected around 2037 (a speculation by me) then we get to the thinking programs and an actual AI. So when we see AI, we need to see that is a program that can course through data and give you the most likely outcome. I will admit that for a lot of people it will fit, but not for all and there we get the problem. You see Microsoft will blame all sources and all kind of people, but in the end it will be up to the programmer to show their algorithm is correct and as I am telling you now that it comes down to unverified data. How does that come over to you?
When you consider that Wired also gave us “it listed numerous GOP candidates who have already pulled out of the race.” The issue of how out of date data is becomes clear. We see all these clever options that others give us, but when some LLM (labeled AI) is un-updated and unreliable, how secure remains your position when you base decision making streams on the wrong data? And that is merely a sales track.
The last teaspoon is given to us by The Guardian. The Guardian (at https://www.theguardian.com/technology/2024/mar/06/microsoft-ai-explicit-image-safety) gave us on March 7th 2024:
3. Microsoft ignored safety problems with AI image generator, engineer complains
So when you consider the previous parts (especially CrowdStrike) “Shane Jones said he warned management about the lack of safeguards several times, but it didn’t result in any action” Microsoft will state that this is another issue. But I spoke about wrong data, out of date data and unverified data. And now we see that the lack of safeguards and inaction would make things worse and a lot faster than you think. You see as long as there is no real AI, all data needs to be verified and that does not seem to be the case in too many setting. I spoke about policy issues and procedural issues. Well here we get the gist “it didn’t result in any action” and we keep on seeing issues with Microsoft. So how many times will you face this? And that is before people realise that their IP are on Azure servers. So how many procedural flaws will your research we driven into until it is all on a Russian or Chinese or North Korean enabled server (most likely by Russia or China, which is a speculation by me).
As such, it was never rocket science, look at any corporation and in their divisions there will always be one person who thinks of number one (himself) and in that setting how safe are you?
There is a reason that I do not want Microsoft near my IP. I can only hope that someone waked up and give me a nice retirement present ($30M post taxation would be nice).
Enjoy the day.
Lawrence van Rijn - Law Lord to be 