Outsourced AI Works Very Well, Thank You
May 2, 2025
Tech experts predict that AI will automate all jobs and make humanity obsolete. If that’s the case then why was so-called AI outsourced? Engadget reports how one “Tech Founder Charged With Fraud For ‘AI’ That Was Secretly Overseas Contract Workers.”
The tech founder in question is Albert Sangier and the US Department of Justice indicated him on misleading clients with Nate, his financial technology platform. Sangier founded Nate in 2018, he raised $40 million from investors, and he claimed that it could give shoppers a universal checkout application powered by AI. The transactions were actually completed by human contractors located in Romania, the Philippines, and bots.
Sangier deception was first noted in 2022:
“ ‘This case follows reporting by The Information in 2022 that cast light on Nate’s use of human labor rather than AI. Sources told the publication that during 2021, “the share of transactions Nate handled manually rather than automatically ranged between 60 percent and 100 percent.’”
Sangier isn’t the only “tech leader” who duplicitously pretends that human workers are actually an AI algorithm or chatbot. More bad actors will do this scam and they’ll get more creative hiding their steps.
Whitney Grace, May 2, 2025
Anthropic Discovers a Moral Code in Its Smart Software
April 30, 2025
No AI. This old dinobaby just plods along, delighted he is old and this craziness will soon be left behind. What about you?
With the United Arab Emirates starting to use smart software to make its laws, the idea that artificial intelligence has a sense of morality is reassuring. Who would want a person judged guilty by a machine to face incarceration, a fine, or — gulp! — worse.
“Anthropic Just Analyzed 700,000 Claude Conversations — And Found Its AI Has a Moral Code of Its Own” explains:
The [Anthropic] study examined 700,000 anonymized conversations, finding that Claude largely upholds the company’s “helpful, honest, harmless” framework while adapting its values to different contexts — from relationship advice to historical analysis. This represents one of the most ambitious attempts to empirically evaluate whether an AI system’s behavior in the wild matches its intended design.
Two philosophers watch as the smart software explains the meaning of “situational and hallucinatory ethics.” Thanks, OpenAI. I bet you are glad those former employees of yours quit. Imagine. Ethics and morality getting in the way of accelerationism.
Plus the company has “hope”, saying:
“Our hope is that this research encourages other AI labs to conduct similar research into their models’ values,” said Saffron Huang, a member of Anthropic’s Societal Impacts team who worked on the study, in an interview with VentureBeat. “Measuring an AI system’s values is core to alignment research and understanding if a model is actually aligned with its training.”
The study is definitely not part of the firm’s marketing campaign. The write up includes this quote from an Anthropic wizard:
The research arrives at a critical moment for Anthropic, which recently launched “Claude Max,” a premium $200 monthly subscription tier aimed at competing with OpenAI’s similar offering. The company has also expanded Claude’s capabilities to include Google Workspace integration and autonomous research functions, positioning it as “a true virtual collaborator” for enterprise users, according to recent announcements.
For $2,400 per year, a user of the smart software would not want to do something improper, immoral, unethical, or just plain bad. I know that humans have some difficulty defining these terms related to human behavior in simple terms. It is a step forward that software has the meanings and can apply them. And for $200 a month one wants good judgment.
Does Claude hallucinate? Is the Anthropic-run study objective? Are the data reproducible?
Hey, no, no, no. What do you expect in the dog-eat-dog world of smart software?
Here’s a statement from the write up that pushes aside my trivial questions:
The study found that Claude generally adheres to Anthropic’s prosocial aspirations, emphasizing values like “user enablement,” “epistemic humility,” and “patient wellbeing” across diverse interactions. However, researchers also discovered troubling instances where Claude expressed values contrary to its training.
Yes, pro-social. That’s a concept definitely relevant to certain prompts sent to Anthropic’s system.
Are the moral predilections consistent?
Of course not. The write up says:
Perhaps most fascinating was the discovery that Claude’s expressed values shift contextually, mirroring human behavior. When users sought relationship guidance, Claude emphasized “healthy boundaries” and “mutual respect.” For historical event analysis, “historical accuracy” took precedence.
Yes, inconsistency depending upon the prompt. Perfect.
Why does this occur? This statement reveals the depth and elegance of the Anthropic research into computer systems whose inner workings are tough for their developers to understand:
Anthropic’s values study builds on the company’s broader efforts to demystify large language models through what it calls “mechanistic interpretability” — essentially reverse-engineering AI systems to understand their inner workings. Last month, Anthropic researchers published groundbreaking work that used what they described as a “microscope” to track Claude’s decision-making processes. The technique revealed counterintuitive behaviors, including Claude planning ahead when composing poetry and using unconventional problem-solving approaches for basic math.
Several observations:
- Unlike Google which is just saying, “We are the leaders,” Anthropic wants to be the good guys, explaining how its smart software is sensitive to squishy human values
- The write up itself is a content marketing gem
- There is scant evidence that the description of the Anthropic “findings” are reliable.
Let’s slap this Anthropic software into an autonomous drone and let it loose. It will be the AI system able to make those subjective decisions. Light it up and launch.
Stephen E Arnold, April 30, 2025
Google Wins AI, According to Google AI
April 29, 2025
No AI. This old dinobaby just plods along, delighted he is old and this craziness will soon be left behind. What about you?
Wow, not even insecure pop stars explain how wonderful they are at every opportunity. But Google is not going to stop explaining that it is number one in smart software. Never mind the lawsuits. Never mind the Deepseek thing. Never mind Sam AI-Man. Never mind angry Googlers who think the company will destroy the world.
Just get the message, “We have won.”
I know this because I read the weird PR interview called “Demis Hassabis Is Preparing for AI’s Endgame,” which is part of the “news” about the Time 100 most wonderful and intelligence and influential and talented and prescient people in the Time world.
Let’s take a quick look at a few of the statements in the marketing story. Because I am a dinobaby, I will wrap up with a few observations designed to make clear the difference between old geezers like me and the youthful new breed of Time leaders.
Here’s the first passage I noted:
He believes AGI [Googler Hassabis] would be a technology that could not only solve existing problems, but also come up with entirely new explanations for the universe. A test for its existence might be whether a system could come up with general relativity with only the information Einstein had access to; or if it could not only solve a longstanding hypothesis in mathematics, but theorize an entirely new one. “I identify myself as a scientist first and foremost,” Hassabis says. “The whole reason I’m doing everything I’ve done in my life is in the pursuit of knowledge and trying to understand the world around us.”
First comment. Yep, I noticed the reference to Einstein. That’s reasonable intellectual territory for a Googler. I want to point out that the Google is in a bit of legal trouble because it did not play fair. But neither did Einstein. Instead of fighting evil in Europe, he lit out for the US of A. I mean a genius of the Einstein ilk is not going to risk one’s life. Just think. Google is a thinking outfit, but I would suggest that its brush with authorities is different from Einstein’s. But a scientist working at an outfit in trouble with authorities, no big deal, right? AI is a way to understand the world around us. Breaking the law? What?
The second snippet is this one:
When DeepMind was acquired by Google in 2014, Hassabis insisted on a contractual firewall: a clause explicitly prohibiting his technology from being used for military applications. It was a red line that reflected his vision of AI as humanity’s scientific savior, not a weapon of war.
Well, that red line was made of erasable market red. It has disappeared. And where is the Nobel prize winner? Still at the Google, that’s the outfit that is in trouble with the law and reasonably good at discarding notions that don’t fit with its goal of generating big revenue from ads and assorted other ventures like self driving taxi cabs. Noble indeed.
Okay, here’s the third comment:
That work [dumping humans for smart software], he says, is not intended to hasten labor disruptions, but instead is about building the necessary scaffolding for the type of AI that he hopes will one day make its own scientific discoveries. Still, as research into these AI “agents” progresses, Hassabis says, expect them to be able to carry out increasingly more complex tasks independently. (An AI agent that can meaningfully automate the job of further AI research, he predicts, is “a few years away.”)
I think that Google will just say, “Yo, dudes, smart software is efficient. Those who lose their jobs can re-skill like the humanoids we are allowing to find their future elsewhere.
Several observations:
- I think that the Time people are trying to balance their fear of smart software replacing outfits like Time with the excitement of watching smart software create a new way experiencing making a life. I don’t think the Timers achieved their goal.
- The message that Google thinks, cares, and has lofty goals just doesn’t ring true. Google is in trouble with the law for a reason. It was smart enough to make money, but it was not smart enough to avoid honking off regulators in some jurisdictions. I can’t reconcile illegal behavior with baloney about the good of mankind.
- Google wants to be seen as the big dog of AI. The problem is that saying something is different from the reality of trials, loss of trust among some customer sectors, floundering for a coherent message about smart software, and the baloney that the quantumly supreme Google convinces people to propagate.
Okay, you may love the Time write up. I am amused, and I think some of the lingo will find its way into the Sundar & Prabhakar Comedy Show. Did you hear the one about Google’s AI not being used for weapons?
Stephen E Arnold, April 29, 2025
China, Self-Amusement, and AI
April 29, 2025
China pokes fun at the United States whenever it can. Why? The Middle Kingdom wants to prove its superiority over the US. China is does have many technological advances over its western neighbor and now the country made another great leap forward with AI says Business Insider: “China’s Baidu Releases Ernie X1, A New AI Reasoning Model.”
Baidu is China’s equivalent of Google and the it released two new AI models. The first is Ernie X1 that is described as a reasoning model that delivers on par with Deepseek R1 at half the price. It also released a multimodal foundation model called Ernie 4.5 that could potentially outperform GPT-4.5 and costs only a fraction of the price. Baidu is also developing the Ernie Bot, a free chatbot.
Baidu wants to offer the world cheap AI:
“Baidu’s new releases come as Silicon Valley reckons with the cost of AI models, largely spurred by the latest drops from Deepseek, a Chinese startup launched by hedge fund High Flyer.
In December, Deepseek released a large language model called V3, and in January, it unveiled a reasoning model called R1. The models are considered as good or better than equivalent models from OpenAI but priced “anywhere from 20-40x cheaper,” according to analysis from Bernstein Research.”
China is smart to develop inexpensive AI, but did the country have to make fun of Sesame Street? I mean Big Bird?
Whitney Grace, April 29, 2025
Synthetic Data: It Sucks, Says Designers
April 25, 2025
Some would argue taking actual customers out of market research is a bad idea. Smashing Magazine supports that reasonable perspective in, “How to Argue Against AI-First Research.” Yes, some AI enthusiasts praise synthetic user testing as a valuable new tool. The practice is exactly what it sounds like—using LLMs to build fake customers and performing market research on them. Admittedly, it is much faster and cheaper than surveying actual humans. But what good is that if the results are bad? Writer Vitaly Friedman explains:
“When ‘producing’ user insights, LLMs can’t generate unexpected things beyond what we’re already asking about. In comparison, researchers are only able to define what’s relevant as the process unfolds. In actual user testing, insights can help shift priorities or radically reimagine the problem we’re trying to solve, as well as potential business outcomes. Real insights come from unexpected behavior, from reading behavioral clues and emotions, from observing a person doing the opposite of what they said. We can’t replicate it with LLMs.”
But budgets are tight. Isn’t synthetic user data better than nothing? No. No it is not. We learn:
“Pavel Samsonov articulates that things that sound like customers might say them are worthless. But things that customers actually have said, done, or experienced carry inherent value (although they could be exaggerated). We just need to interpret them correctly. AI user research isn’t ‘better than nothing’ or ‘more effective.’ It creates an illusion of customer experiences that never happened and are at best good guesses but at worst misleading and non-applicable.”
Not only that, cutting real customers out of the equation means not catching AI errors. And there will be errors. Furthermore, emphasizes Friedman:
“Synthetic testing assumes that people fit in well-defined boxes, which is rarely true. Human behavior is shaped by our experiences, situations, habits that can’t be replicated by text generation alone. AI strengthens biases, supports hunches, and amplifies stereotypes.”
All of which could send marketing dollars down the wrong, unprofitable track. As suspicious as we are of AI hype, even we can admit the tech is good for some things. Market research perhaps is not a core competency.
Cynthia Murrell, April 25, 2025
AI Crawlers Are Bullying Open Source: Stop Grousing and Go Away
April 25, 2025
AI algorithms are built on open source technology. Unfortunately generative AI is harming its mother code explains TechDirt: “AI Crawlers Are Harming Wikimedia, Bringing Open Source Sites To Their Knees, And Putting The Open Web At Risk.” To make generative AI work you need a lot of computer power, smart coding, and mounds of training data. Money can buy coding and power, but (quality) training data is incredibly difficult to obtain.
AI crawlers were unleashed on the Internet to scrap information and use it for training models. The biggest information providers for crawlers are Wikimedia projects and it’s a big problem. Wikimedia, which claims to be “the largest collection of open knowledge in the world,” says most of its traffic is from crawlers and it is eating into costs:
“Since January 2024, we have seen the bandwidth used for downloading multimedia content grow by 50%. This increase is not coming from human readers, but largely from automated programs that scrape the Wikimedia Commons image catalog of openly licensed images to feed images to AI models. Our infrastructure is built to sustain sudden traffic spikes from humans during high-interest events, but the amount of traffic generated by scraper bots is unprecedented and presents growing risks and costs.”
This is bad because it is straining the Wikimedia datacenter and budgetary resources. Wikimedia isn’t the only information source feeling the burn from AI crawlers. News sites and more are being wrung by crawlers for every decimal of information:
“It’s increasingly clear that the reckless and selfish way in which AI crawlers are being deployed by companies eager to tap into today’s AI hype is bringing many sites around the Internet to their knees. As a result, AI crawlers are beginning to threaten the open Web itself, and thus the frictionless access to knowledge that it has provided to general users for the last 30 years.”
Silicon Valley might have good intentions but dollars are more important. (Oh, I am not sure about the “good intentions.”)
Whitney Grace, April 25, 2025
Microsoft and Its Modern Management Method: Waffling
April 23, 2025
No AI, just the dinobaby himself.
The Harvard Business School (which I assume will remain open for “business”) has not addressed its case writers to focus on Microsoft’s modern management method. To me, changing direction is not a pivot; it is a variant of waffling. “Waffling” means saying one thing like “We love OpenAI.” Then hiring people who don’t love OpenAI and cutting deals with other AI outfits. The whipped cream on the waffle is killing off investments in data centers.
If you are not following this, think of the old song “The first time is the last time,” and you might get a sense of the confusion that results from changes in strategic and tactical direction. You may find this GenX, Y and Z approach just fine. I think it is a hoot.
PC Gamer, definitely not the Harvard Business Review, tackles one example of Microsoft’s waffling in “Microsoft Pulls Out of Two Big Data Centre Deals Because It Reportedly Doesn’t Want to Support More OpenAI Training Workloads.”
The write up says:
Microsoft has pulled out of deals to lease its data centres for additional training of OpenAI’s language model ChatGPT. This news seems surprising given the perceived popularity of the model, but the field of AI technology is a contentious one, for a lot of good reasons. The combination of high running cost, relatively low returns, and increasing competition—plus working on it’s own sickening AI-made Quake 2 demo—have proven enough reason for Microsoft to bow out of two gigawatt worth of projects across the US and Europe.
I love the scholarly “sickening.” Listen up, HBR editors. That’s a management term for 2025.
The article adds:
Microsoft, as well as its investors, have witnessed this relatively slow payoff alongside the rise of competitor models such as China’s Deepseek.
Yep, “payoff.” The Harvard Business School’s professors are probably not familiar with the concept of a payoff.
The news report points out that Microsoft is definitely, 100 percent going to spend $80 billion on infrastructure in 2025. With eight months left in the year, the Softies have to get in gear. The Google is spending as well. The other big time high tech AI juggernauts are also spending.
Will these investments payoff? Sure. Accountants and chief financial officers learn how to perform number magic. Guess where? Schools like the HBS. Don’t waffle. Go to class. Learn and then implement big time waffling.
Stephen E Arnold, April 23, 2025
ArXiv: Will Other Smart Software Systems Get “Free” Access? Yeah, Sure
April 21, 2025
Believe it or not, no smart software. Just a dumb and skeptical dinobaby.
Before commenting on Cornell University’s apparent shift of the ArXiv service to the Google Cloud, let me point you to this page:
The page was updated 15 years ago. Now check out the access to
NCSTRL, the Networked Computer Science Technical Reference Library.
CoRR, the Computing Research Repository.
The Open Archives Initiative.
ETRDL, the ERCIM Technical Reference Digital Library.
Cornell University Library Historical Math Book Collection
Cornell University Library Making of America Collection
Hein online Retrospective Law Journals
Yep, 404s, some content behind paywalls, and other data just disappeared because Bing, Google, and Yandex don’t index certain information no matter what people believe or the marketers say.
This orphaned Cornell University Dienst service has “gorged out”; that is, jumped off a bridge to the rocks below. The act is something students know about but the admissions department seems to not be aware of the bound phrase.
I read “Careers at ArXiv.” The post seems to say to me, “We are moving the ArXiv “gray” papers to Google Cloud. Here’s a snippet of the “career” advertisement / news announcement:
We are already underway on the arXiv CE ("Cloud Edition") project. This is a project to re-home all arXiv services from VMs at Cornell to a cloud provider (Google Cloud). There are a number of reasons for this transition, including improving arXiv’s scalability while modernizing our infrastructure. This will not be a simple port of the existing arXiv code base because this project will:
- replace the portion of our backends still written in perl and PHP
- re-architect our article processing to be fully asynchronous, and provide better insight into the processing workflows
- containerize all, or nearly all arXiv services so we can deploy via Kubernetes or services like Google Cloud Run
- improve our monitoring and logging facilities so we can more quickly identify and manage production issues with arxiv.org
- create a robust CI/CD pipeline to give us more confidence that changes we deploy will not cause services to regress
The cloud transition is a pre-requisite to modernizing arXiv as a service. The modernization will enable: – arXiv to expand the subject areas that we cover – improve the metadata we collect and make available for articles, adding fields that the research community has requested such as funder identification – deal with the problem of ambiguous author identities – improve accessibility to support users with impairments, particularly visual impairments – improve usability for the entire arXiv community.
I know Google is into “free.” The company is giving college students its quantumly supreme smart software for absolutely nothing. Maybe a Google account will be required? Maybe the Chrome browser may be needed to give those knowledge hungry college students the best experience possible? Maybe Google’s beacons, bugs, and cookies will be the students’ constant companions? Yeah, maybe.
But will ArXiv exist in the future? Will Google’s hungry knowledge munchers chew through the data and then pull a Dienst maneuver?
As a dinobaby, I liked the ArXiv service, but I also liked the Dienst math repository before it became unfindable.
It seems to me that Cornell University is:
- Saving money at the library and maybe the Theory Center
- Avoiding future legal dust ups about access to content which to some government professionals may reveal information to America’s adversaries
- Intentionally or inadvertently giving the Google control over knowledge flow related to matters of technical and competitive interest to everyone’s favorite online advertising company
- Running a variation of its Dienst game plan.
But I am a dinobaby, and I know zero about Cornell other than the “gorging out” approach to termination. I know even less about the blue chip consulting type thinking in which the Google engages. I don’t even know if I agree that Google’s recent court loss is really a “win” for the Google.
But the future of the ArXiv? Hey, where is that bridge? Do some students jump, fall, or get pushed to their death on the rocks below?
PS. In case your German is rusty “dienst” means duty and possibly “a position of authority” like a leader at Google.
Stephen E Arnold, April xx, 2025
AI and Movies: Better and Cheaper!
April 21, 2025
Believe it or not, no smart software. Just a dumb and skeptical dinobaby.
I am not a movie oriented dinobaby. I do see occasional stories about the motion picture industry. My knowledge is shallow, but several things seem to be stuck in my mind:
- Today’s movies are not too good
- Today’s big budget films are recycles of sequels, pre-quels, and less than equals
- Today’s blockbusters are expensive.
I did a project for a little-time B movie fellow. I have even been to an LA party held in a mansion in La Jolla. I sat in the corner in my brown suit and waited until I could make my escape.
End of Hollywood knowledge.
I read “Ted Sarandos Responds To James Cameron’s Vision Of AI Making Movies Cheaper: “There’s An Even Bigger Opportunity To Make Movies 10% Better.” No, I really did red the article. I cam away confused. Most of my pre-retirement work involved projects whose goal was to make a lot of money. The idea was be clever, do a minimum of “real” work, and then fix up the problems when people complained. The magic formula for some Silicon Valley and high-technology outfits located outside of the Plastic Fantastic World.
This article pits better versus cheaper. I learned:
Citing recent comments by James Cameron, Netflix Co-CEO Ted Sarandos said he hopes AI can make films “10% better,” not just “50% cheaper.”
Well, there you go. Better and cheaper. Is that the winning formula for creative work? The write up quotes Ted Sarandos (a movie expert, I assume) as saying:
Today, you can use these AI-powered tools to enable smaller-budget projects to have access to big VFX on screen.
From my point of view “better” means more VFX which is, I assume, movie talk for visual effects. These are the everyday things I see at my local grocery store. There are super heroes stopping crimes in progress. There are giant alien creatures shooting energy beams at military personnel. There are machines that have a great voice that some AI experts found particularly enchanting.
The cheaper means that the individuals who sit in front of computer screens fooling around with Blackmagic’s Fusion and the super-wonderful Adobe software will be able to let smart software do some of the work. If 100 people work on a big budget film’s VFX and smart software can do the work cheaper, the question arises, “Do we need these 100 people?” Based on my training, the answer is, “Nope. Let them find their future elsewhere.”
The article sidesteps two important questions: Question 1. What does better mean? Question 2. What does cheaper mean?
Better is subjective. Cheaper is a victim of scope creep. Big jobs don’t get cheaper. Big jobs get more expensive.
What smart software will do the motion picture industry is hasten its “re-invention.”
The new video stars are those who attract eyeballs on TikTok- and YouTube-type platforms. The traditional motion picture industry which created yesterday’s stars or “influencers” is long gone. AI is going to do three things:
- Replace skilled technicians with software
- Allow today’s “influencers” to become the next Clark Gabel and Marilyn Monroe (How did she die?)
- Reduce the barrier for innovations that do not come from recycling Superman-type pre-quels, sequels, and less than equals.
To sum up, both of these movie experts are right and wrong. I suppose both can be reskilled. Does Mr. Beast offer a for fee class on video innovation which includes cheaper production and better outputs?
Stephen E Arnold, April 21, 2025
When
Bugs, Debugs, and Rebugs: AI Does Great Work
April 21, 2025
AI algorithms have already been assimilated into everyday technology, but AI still has problems or you could say there’s a bug in their code. TechCrunch tells us that, “AI Models Still Struggle To Debug Software, Microsoft Study Shows.” Large AI Models such as Anthropic, OpenAI, and more are used for programming. Mark Zuckerberg of Facebook plans to deploy AI coding models at his company, while Sundar Pichai, the Google CEO, said that 25% of code is AI generated.
AI algorithms are great at automating tasks, but they shouldn’t be relied on 100% for all programming projects. Microsoft Research released a new study that discovered AI models like Cause 3.7 Sonnet and 03-mini fail to debug problems in SWE-bench Lite, a software development benchmark. Humans still beat technology when it comes to coding. Here’s what the study did and found:
“The study’s co-authors tested nine different models as the backbone for a “single prompt-based agent” that had access to a number of debugging tools, including a Python debugger. They tasked this agent with solving a curated set of 300 software debugging tasks from SWE-bench Lite. According to the co-authors, even when equipped with stronger and more recent models, their agent rarely completed more than half of the debugging tasks successfully. Claude 3.7 Sonnet had the highest average success rate (48.4%), followed by OpenAI’s o1 (30.2%), and o3-mini (22.1%).”
What is the problem? It’s one that AI has faced since it was first programmed: lack of data for training.
More studies show that AI generated code creates security vulnerabilities too. Is anyone surprised? (Just the AI marketers who do not understand why their assertions don’t match reality.)
Whitney Grace, April 21, 2025
-
- Subscribe to Beyond Search
Feature archive
News archive
-
