AI, Math, and Cognitive Dissonance
July 28, 2025
This blog post is the work of an authentic dinobaby. Sorry. No smart software can help this reptilian thinker.
AI marketers will have to spend some time positioning their smart software as great tools for solving mathematical problems. “Not Even Bronze: Evaluating LLMs on 2025 International Math Olympiad” reports that words about prowess are disconnected from performance. The write up says:
The best-performing model is Gemini 2.5 Pro, achieving a score of 31% (13 points), which is well below the 19/42 score necessary for a bronze medal. Other models lagged significantly behind, with Grok-4 and Deepseek-R1 in particular underperforming relative to their earlier results on other MathArena benchmarks.
The write up points out, possibly to call attention to the slight disconnect between the marketing of Google AI and its performance in this contest:
As mentioned above, Gemini 2.5 Pro achieved the highest score with an average of 31% (13 points). While this may seem low, especially considering the $400 spent on generating just 24 answers, it nonetheless represents a strong performance given the extreme difficulty of the IMO. However, these 13 points are not enough for a bronze medal (19/42). In contrast, other models trail significantly behind and we can already safely say that none of them will achieve the bronze medal. Full results are available on our leaderboard, where everyone can explore and analyze individual responses and judge feedback in detail.
This is one “competition”, the lousy performance of the high-profile models, and the complex process required to assess performance make it easy to ignore this result.
Let’s just assume that it is close enough for horse shoes and good enough. With that assumption in mind, do you want smart software making decisions about what information you can access, the medical prognosis for your nine-year-old child, or decisions about your driver’s license renewal?
Now, let’s consider this write up fragmented across Tweets: [Thread] An OpenAI researcher says the company’s latest experimental reasoning LLM achieved gold medal-level performance on the 2025 International Math Olympiad. The little posts are perfect for a person familliar with TikTok-type and Twitter-like content. Not me. The main idea is that in the same competition, OpenAI earned “gold medal-level performance.”
The $64 dollar question is, “Who is correct?” The answer is, “It depends.”
Is this an example of what I learned in 1962 in my freshman year at a so-so university? I think the term was cognitive dissonance.
Stephen E Arnold, July 28, 2025
Silicon Valley: The New Home of Unsportsmanlike Conduct
July 26, 2025
Sorry, no smart software involved. A dinobaby’s own emergent thoughts.
I read the Axios run down of Mark Zuckerberg’s hiring blitz. “Mark Zuckerberg Details Meta’s Superintelligence Plans” reports:
The company [Mark Zuckerberg’s very own Meta] is spending billions of dollars to hire key employees as it looks to jumpstart its effort and compete with Google, OpenAI and others.
Meta (formerly the estimable juicy brand Facebook) had some smart software people. (Does anyone remember Jerome Pesenti?) Then there was Llama, and like the guanaco, tamed and used to carry tourists to Peruvian sights, has been seen as a photo opp for parents wanting to document their kids’ visit to Cusco.
Is Mr. Zuckerberg creating a mini Bell Labs in order to take the lead in smart software?The Axios write up contains some names of people who may have some connection to the Middle Kingdom. The idea is to get smart people, put them in a two-story building in Silicon Valley, turn up the A/C, and inject snacks.
I interpret the hiring and the allegedly massive pay packets to a simpler, more direct idea: Move fast, break things.
What are the things Mr. Zuckerberg is breaking?
First, I worked in Silicon Valley (aka Plastic Fantastic) for a number of years. I lived in Berkely and loved that commute to San Mateo, Foster City, and environs. Poaching employees was done in a more relaxed way. A chat at a conference, a small gathering after a softball game at the public fields not far from Stanford (yes, the school which had a president who made up information), or at some event like a talk at the Computer Museum or whatever it was called. That’s history. Mr. Zuckerberg shows up (virtually or in a T shirt), offers an alleged $100 million and hires a big name. No muss. No fuss. No social conventions. Just money. Cash. (I almost wish I was 25 and working in Mountain View. Sigh.)
Second, Mr. Zuckerberg is targeting the sensitive private parts of big leadership people. No dancing. Just targeted castration of key talent. Ouch. The Axios write up provides the names of some of these individuals. What interesting is that these people come from the knowledge parts hidden from the journalistic spotlight. Those suffering life changing removals without anesthesia include Google, OpenAI, and similar firms. In the good old days, Silicon Valley firms competed less of that Manhattan, lower east side vibe. No more.
Third, Mr. Zuckerberg is not announcing anything at conferences or with friendly emails. He is just taking action. Let the people at Apple, Safe Superintelligence, and similar outfits read the news in a resignation email. Mr. Zuckerberg knows that those NDAs and employment contracts can be used to wipe away tears when the loss of a valuable person is discovered.
What’s up?
Obviously Mr. Zuckerberg is not happy that his outfit is perceived as a loser in the AI game. Will this Bell Labs’ West approach work? Probably not. It will deliver one thing, however. Mr. Zuckerberg is sending a message that he will spend money to cripple, hobble, and derail AI innovation at firms beating his former LLM to death.
Move fast and break things has come to the folks who used the approach to take out swaths of established businesses. Now the technique is being used on companies next door. Welcome to the ungentrified neighborhood. Oh, expect more fist fights at those once friendly, co-ed softball games.
Stephen E Arnold, July 26, 2025
Decentralization: Nope, a Fantasy It Seems
July 25, 2025
Just a dinobaby working the old-fashioned way, no smart software.
Web 3, decentralization, graceful fail over, alternative routing. Are these concepts baloney? I think the idea that the distributed approach to online systems is definitely not bullet proof.
Why would I, an online person, make such a statement? I read “Cloudflare 1.1.1.1 Incident on July 14, 2025.” I know a number of people who know zero about Cloudflare. One can argue that AT&T, Google, Microsoft, et al are the gate keepers of the online world. Okay, that sounds great. It is sort of true.
I quote from the write up:
For many users, not being able to resolve names using the 1.1.1.1 Resolver meant that basically all Internet services were unavailable.
The operative word is “all.”
What can one conclude if this explanation of a failure of “legacy” systems can be pinned on a “configuration error.”? Some observations:
- A bad actor able to replicate this can kill the Internet or at least Cloudflare’s functionality
- The baloney about decentralization is just that… baloney. Cheap words packed in a PR tube and “sold” as something good
- The fail over and resilience assertions? Three-day old fish. Remember Ben Franklin’s aphorism: Three-day old fish smell. Badly.
Net net: We have evidence that the reality of today’s Internet rests in the semi capable hands of certain large companies. Without real “innovation,” the centralization of certain functions will have wide spread and unexpected impacts. Yep, “all,” including the bad actors who make use of these points of concentration. The Cloudflare incident may motivate other technically adept groups to find a better way. Perhaps something in the sky like satellites or on the ground like device to device wireless? I wonder if adversaries of the US have noticed this incident?
Stephen E Arnold, July 25, 2025
Will Apple Do AI in China? Subsidies, Investment, Saluting Too
July 25, 2025
![Dino 5 18 25_thumb[3]](https://www.arnoldit.com/wordpress/wp-content/uploads/2025/07/Dino-5-18-25_thumb3-2.gif "Dino 5 18 25_thumb[3]")
This blog post is the work of an authentic dinobaby. Sorry. No smart software can help this reptilian thinker.
Apple long ago vowed to use the latest tech to design its hardware. Now that means generative AI. Asia Financial reports, “Apple Keen to Use AI to Design Its Chips, Tech Executive Says.” That tidbit comes from a speech Apple VP Johny Srouji made as he accepted an award from tech R&D group Imec. We learn:
“In the speech, a recording of which was reviewed by Reuters, Srouji outlined Apple’s development of custom chips from the first A4 chip in an iPhone in 2010 to the most recent chips that power Mac desktop computers and the Vision Pro headset. He said one of the key lessons Apple learned was that it needed to use the most cutting-edge tools available to design its chips, including the latest chip design software from electronic design automation (EDA) firms. The two biggest players in that industry – Cadence Design Systems and Synopsys – have been racing to add artificial intelligence to their offerings. ‘EDA companies are super critical in supporting our chip design complexities,’ Srouji said in his remarks. ‘Generative AI techniques have a high potential in getting more design work in less time, and it can be a huge productivity boost.’”
Srouji also noted Apple is one to commit to its choices. The post notes:
“Srouji said another key lesson Apple learned in designing its own chips was to make big bets and not look back. When Apple transitioned its Mac computers – its oldest active product line – from Intel chips to its own chips in 2020, it made no contingency plans in case the switch did not work.”
Yes, that gamble paid off for the polished tech giant. Will this bet be equally advantageous?
Has Apple read “Apple in China”?
Cynthia Murrell, July 25, 2025
Microsoft, Security, and Blame: Playing the Same Record Again
July 24, 2025
This blog post is the work of an authentic dinobaby. Sorry. No smart software can help this reptilian thinker.
I have a dim memory of being at my friend’s house. His sister was playing “Catch a Falling Star” by Perry Como. My friend’s mother screamed, “Turn off that music. It’s driving me crazy.” The repetition, the loudness, and the sappiness were driving my friend’s mother insane. I didn’t care. I have the ability to tune out repetition, loud noise, and sappiness. My friend’s sister turned up the record player. Words did not work.
Those skills were required when I read “Microsoft Says Chinese Hackers Are Exploiting SharePoint Flaws.” The write up reports:
Microsoft Corp. accused Chinese hackers of exploiting vulnerabilities in its SharePoint software that have led to breaches worldwide in recent days.
What does one expect? Microsoft has marketed its software to government agencies and companies throughout the world. Hundreds of millions of people use its products and services. Students in computer science security classes learn how to probe its ubiquitous software for weak points. Professionals exploit hunters target software in wide use.
When a breach occurs, what tune does Microsoft put on the record player? The song is “Blame Game.” One verse is:
Let’s play the blame game, I love you more
Let’s play the blame game for sure
Let’s call out names, names, I hate you more
Let’s call out names, names, for sure
My dinobaby thought is that the source of the problem is not Chinese bad actors or thousands of Russian hackers or whatever assertion is presented to provide cover for a security failure.
Why not address the issue of Microsoft’s own quality control processes? Whatever happened to making security Job One? Oh, right, AI is the big deal. Well, if the AI is so good, why doesn’t Microsoft’s AI address these issues directly.
Maybe Microsoft is better at marketing than at software engineering? Now that’s a question worth exploring at the US government agencies now at risk of Microsoft’s own work processes.
Mothers can shout at their children. Microsoft issues PR speak about government intelligence agencies. News flash, Microsoft. Those actors know what big soft target to attack. Plus, they are not listening to your old tunes.
Stephen E Arnold, July 24, 2024
AI Content Marketing: Claims about Savings Are Pipe Dreams
July 24, 2025
This blog post is the work of an authentic dinobaby. Sorry. No smart software can help this reptilian thinker.
My tiny team and I sign up for interesting smart software “innovations.” We plopped down $40 to access 1min.ai. Some alarm bells went off. These were not the panic inducing Code Red buzzers at the Google. But we noticed. First, registration was wonky. After several attempts were had an opportunity to log in. After several tries, we gained access to the cornucopia of smart software goodies. We ran one query and were surprised to see Hamster Kombat-style points. However, the 1min.ai crowd flipped the winning click-to-earn model on its head. Every click consumed points. When the points were gone, the user had to buy more. This is an interesting variation of taxi meter pricing, a method reviled in the 1980s when commercial databases were the rage.
I thought about my team’s experience with 1min.ai and figured that an objective person would present some of these wobbles. Was I wrong? Yes.
“Your New AI-Powered Team Costs Less Than $80. Meet 1min.ai” is one of the wildest advertorial or content marketing smoke screens I have encountered in the last week or so. The write up asserts as actual factual, hard-hitting, old-fashioned technology reporting:
If ChatGPT’s your sidekick, think of 1min.AI as your entire productivity squad. This AI-powered tool lets you automate all kinds of content and business tasks—including emails, social media posts, blog drafts, reports, and even ad copy—without ever opening a blank doc.
I would suggest that one might tap 1min.ai to write an article for a hard-working, logic-charged professional at Macworld.
How about this descriptive paragraph which may have been written by an entity or construct:
Built for speed and scale, 1min.AI gives you access to over 80 AI tools designed to handle everything from content generation to data analysis, customer support replies, and more. You can even build your own tools inside the platform using its AI builder—no coding required.
And what about this statement:
The UI is slick and works in any browser on macOS.
What’s going on?
First, this information is PR assertions without factual substance.
Two, the author did not try to explain the taxi meter business model. It is important if one uses one account for a “team.”
Three, the functionality of the system is less useful that You.com based on our tests. Comparing 1min.ai is a key word play. ChatGPT has some bit time flaws. These include system crashes and delivering totally incorrect information. But 1min.ai lags behind. When ChatGPT stumbles over the prompt finish line, 1min.ai is still lacing its sneakers.
Here’s the final line of this online advertorial:
Act now while plans are still in stock!
How does a digital subscription go out of stock. Isn’t the offer removed?
I think more of this type of AI play acting will appear in the months ahead.
Stephen E Arnold, July 24, 2025
AI and Customer Support: Cost Savings, Yes. Useful, No
July 24, 2025
This blog post is the work of an authentic dinobaby. Sorry. No smart software can help this reptilian thinker.
AI tools help workers to be more efficient and effective, right? Not so much. Not in this call center, anyway. TechSpot reveals, “Call Center Workers Say Their AI Assistants Create More Problems than They Solve.” How can AI create problems? Sure, it hallucinates and it is unpredictable. But why should companies let that stop them? They paid a lot for these gimmicks, after all.
Writer Rob Thubron cites a study showing customer service reps at a Chinese power company are less than pleased with their AI assistants. For one thing, the tool often misunderstands customers’ accents and speech patterns, introducing errors into call transcripts. Homophones are a challenge. It also struggles to accurately convert number sequences to text—resulting in inaccurate phone numbers and other numeric data.
The AI designers somehow thought their product would be better at identifying human emotions than people. We learn:
“Emotion recognition technology, something we’ve seen several reports about – most of them not good – is also criticized by those interviewed. It often misclassified normal speech as being a negative emotion, had too few categories for the range of emotions people expressed, and often associated a high volume level as someone being angry or upset, even if it was just a person who naturally talks loudly. As a result, most CSRs [Customer Service Reps] ignored the emotional tags that the system assigned to callers, saying they were able to understand a caller’s tone and emotions themselves.”
What a surprise. Thubron summarizes:
“Ultimately, while the AI assistant did reduce the amount of basic typing required by CSRs, the content it produced was often filled with errors and redundancies. This required workers to go through the call summaries, correcting mistakes and deleting sections. Moreover, the AI often failed to record key information from customers.”
Isn’t customer service rep one of the jobs most vulnerable to AI takeover? Perhaps not, anymore. A June survey from Gartner found half the organizations that planned to replace human customer service reps with AI are doing an about-face. A couple weeks later, the research firm anticipated that more than 40% of agentic AI projects will be canceled by 2027. Are the remaining 60% firms that have sunk too much money into such ventures to turn back?
Cynthia Murrell, July 24, 2025
The Difference between Financials Viewed by a PR Person and an MBA
July 23, 2025
This blog post is the work of an authentic dinobaby. Sorry. No smart software can help this reptilian thinker.
Another essay from a public relations professional. I just wrote a short blog post about “Clarity.” Now I am addressing an essay titled “The Hater’s Guide To The AI Bubble.” The write up is a fairly long one, about 14,000 or so. The length makes clear that considerable work by the author or assistants went into compiling the numeric data about investments by big outfits in artificial intelligence.
The John Milton essay Areopagitica contains about 16,000 words. This puts the cited article squarely in the readership target of an educated readership in 1644. Like Milton’s screed, “The Hater’s Guide to the AI Bubble” is a polemic.
I am not going to create an abstract of the write up. Instead I will feed the text into some smart software and ask for a 150 word summary just like the ones in the old-fashioned ABI/INFORM database. Here’s what the cheap version of ChatGPT output:
Edward Zitron critiques the generative AI boom, arguing it is an unsustainable financial and technological bubble driven by hype rather than profitability or utility. He contends that companies like Microsoft, Google, and Meta have spent over $560 billion on AI infrastructure with little to no return, while only NVIDIA profits from the GPU demand. Zitron dismantles comparisons to past innovations like AWS, noting that generative AI lacks infrastructure value, scalability, and viable business models. He criticizes AI “agents” as misleading marketing for underperforming chatbots and highlights that nearly all AI startups are unprofitable. The illusion of widespread AI adoption is, according to Zitron, a coordinated market fantasy supported by misleading press and executive spin. The industry’s fragility, he warns, rests entirely on continued GPU sales. Zitron concludes with a call for accountability, asserting that the current AI trade endangers economic stability and reflects a failure of both corporate vision and journalistic scrutiny. (Source: ChatGPT, cheap subscription, July 22, 2025)
I will assume that you, as I did, worked through the essay. You have firmly in mind that large technology outfits have a presumed choke-hold on smart software. The financial performance of the American high technology sector needs smart software to be “the next big thing.” My view is that offering negative views of the “big thing” are likely to be greeted with the same negative attitudes.
Consider John Milton, blind, assisted by a fellow who visualized peaches at female anatomy, working on a Latinate argument against censorship. He published Areopagitica as a pamphlet and no one cared in 1644. Screeds don’t lead. If something bleeds, it gets the eyeballs.
My view of the write up is:
- PR expert analysis of numbers is different from MBA expert analysis of numbers. The gulf, as validated by the Hater’s Guide, is wide and deep
- PR professionals will not make AI succeed or fail. This is not a Dog the Bounty Hunter type of event. The palpable need to make probabilistic, hallucinating software “work” is truly important, not just to the companies burning cash in the AI crucibles, but to the US itself. AI is important.
- The fear of failure is creating a need to shovel more resources into the infrastructure and code of smart software. Haters may argue that the effort is not delivering; believers have too much skin in the game to quit. Not much shames the tech bros, but failure comes pretty close to making these wizards realize that they too put on pants the same way as other people do.
Net net: The cited write up is important as an example of 21st-century polemicism. Will Mr. Zuckerberg stop paying millions of dollars to import AI talent from China? Will evaluators of the AI systems deliver objective results? Will a big-time venture firm with a massive investment in AI say, “AI is a flop”?
The answer to these questions is, “No.”
AI is here. Whether it is any good or not is irrelevant. Too much money has been invested to face reality. PR professionals can do this; those people writing checks for AI are going to just go forward. Failure is not an option. Talking about failure is not an option. Thinking about failure is not an option.
Thus, there is a difference between how a PR professional and an MBA professional views the AI spending. Never the twain shall meet.
As Milton said in Areopagitica :
“A man may be a heretic in the truth; and if he believes things only because his pastor says so, or the assembly so determines, without knowing other reason, though his belief be true, yet the very truth he holds becomes his heresy. There is not any burden that some would gladlier post off to another, than the charge and care of their religion.”
And the religion for AI is money.
Stephen E Arnold, July 23, 2025
Mixed Messages about AI: Why?
July 23, 2025
Just a dinobaby working the old-fashioned way, no smart software.
I learned that Meta is going to spend hundreds of billions for smart software. I assume that selling ads to Facebook users will pay the bill.
If one pokes around, articles like “Enterprise Tech Executives Cool on the Value of AI” turn up. This write up in BetaNews says:
The research from Akkodis, looking at the views of 500 global Chief Technology Officers (CTOs) among a wider group of 2,000 executives, finds that overall C-suite confidence in AI strategy dropped from 69 percent in 2024 to just 58 percent in 2025. The sharpest declines are reported by CTOs and CEOs, down 20 and 33 percentage points respectively. CTOs also point to a leadership gap in AI understanding, with only 55 percent believing their executive teams have the fluency needed to fully grasp the risks and opportunities associated with AI adoption. Among employees, that figure falls to 46 percent, signaling a wider AI trust gap that could hinder successful AI implementation and long-term success.
Okay. I know that smart software can do certain things with reasonable reliability. However, when I look for information, I do my own data gathering. I think pluck items which seem useful to me. Then I push these into smart AI services and ask for error identification and information “color.”
The result is that I have more work to do, but I would estimate that I find one or two useful items or comments out of five smart software systems to which I subscribe.
Is that good or bad? I think that for my purpose, smart software is okay. However, I don’t ask a question unless I have an answer. I want to get additional inputs or commentary. I am not going to ask a smart software system a question to which I do not think I know the answer. Sorry. My trust in high-flying Google-type Silicon Valley outfits is non existent.
The write up points out:
The report also highlights that human skills are key to AI success. Although technical skill are vital, with 51 percent of CTOs citing specialist IT skills as the top capability gap, other abilities are important too, including creativity (44 percent), leadership (39 percent) and critical thinking (36 percent). These skills are increasingly useful for interpreting AI outputs, driving innovation and adapting AI systems to diverse business contexts.
I don’t agree with the weasel word “useful.” Knowing the answer before firing off a prompt is absolutely essential.
Thus, we have a potential problem. If the smart software crowd can get people who do not know the answers to questions, these individuals will provide the boost necessary to keep this technical balão de fogo up in the air. If not, up in flames.
Stephen E Arnold, July 23, 2025
A Security Issue? What Security Issue? Security? It Is Just a Normal Business Process.
July 23, 2025
Just a dinobaby working the old-fashioned way, no smart software.
I zipped through a write up called “A Little-Known Microsoft Program Could Expose the Defense Department to Chinese Hackers.” The word program does not refer to Teams or Word, but to a business process. If you are into government procurement, contractor oversight, and the exiting world of inspector generals, you will want to read the 4000 word plus write up.
Here’s a passage I found interesting:
Microsoft is using engineers in China to help maintain the Defense Department’s computer systems — with minimal supervision by U.S. personnel — leaving some of the nation’s most sensitive data vulnerable to hacking from its leading cyber adversary…
The balance of the cited article explain what’s is going on with a business process implemented by Microsoft as part of a government contract. There are lots of quotes, insider jargon like “digital escort,” and suggestions that the whole approach is — how can I summarize it? — ill advised, maybe stupid.
Several observations:
- Someone should purchase a couple of hundred copies of Apple in China by Patrick McGee, make it required reading, and then hold some informal discussions. These can be modeled on what happens in the seventh grade; for example, “What did you learn about China’s approach to information gathering?”
- A hollowed out government creates a dependence on third-parties. These vendorsdo not explain how outsourcing works. Thus, mismatches exist between government executives’ assumptions and how the reality of third-party contractors fulfill the contract.
- Weaknesses in procurement, oversight, continuous monitoring by auditors encourage short cuts. These are not issues that have arisen in the last day or so. These are institutional and vendor procedures that have existed for decades.
Net net: My view is that some problems are simply not easily resolved. It is interesting to read about security lapses caused by back office and legal processes.
Stephen E Arnold, July 23, 2025
-
- Subscribe to Beyond Search
Feature archive
News archive
-
