From $20 a Month to $20K a Month. Great Idea… or Not?
March 10, 2025
Another post from the dinobaby. Alas, no smart software used for this essay.
OpenAI was one of many smart software companies. If you meet the people on my team, you will learn that I dismissed most of the outfits as search-and-retrieval outfits looking for an edge. Search definitely needs an edge, but I was not confident that predictive generation of an “answer” was a solution. It was a nifty party trick, but then the money started flowing. In January 2023, Microsoft put Google’s cute sharp teeth on edge. Suddenly AI or smart software was the next big thing. The virtual reality thing did not ring the bell. The increasingly weird fiddling with mobile phones did not get the brass ring. And the idea of Apple becoming the next big thing in chips has left everyone confused. My M1 devices work pretty well, and unless I look at the label on the gizmos I can tell an M1 from and M3. Do I care? Nope.
But OpenAI became news. It squabbled with the mastermind of “renewable” satellites, definitely weird trucks, and digging tunnels in Las Vegas. (Yeah, nice idea, just not for anyone who does not want to get stalled in traffic.) When ChatGPT became available, one of those laboring in my digital vineyards signed me up. I fiddled with it and decided that I would run some of my research through the system. I learned that my research was not in the OpenAI “system.” I had it do some images. Those sucked. I will cancel this week.
I put in my AI folder this article “OpenAI’s is Getting Ready to Release PhD Level AI Agents.” I was engaging in some winnowing and I scanned it. In early February 2025, Digital Marketing News wrote about PhD level agents. I am not a PhD. I quite before I finished my dissertation to work in the really socially conscious nuclear unit of that lovable outfit Halliburton. You know the company. That’s the one that charged about $950.00 for a gallon of fuel during the Iraq war. You will also associate Dick Cheney, a fun person, with the company. So no PhD for me.
I was skeptical because of the dismal performance of ChatGPT 4, oh, whatever, trying to come up with the information I have assembled for my new book for law enforcement professionals. Then I read a Slashdot post with the title “OpenAI Plots Charging $20,000 a Month For PhD-Level Agents” shared from a publication I don’t know much about. I think it is like 404 or a for-fee Substack. The publication has great content, and you have to pay for it.
Be that as it may, the Slashdot post reports or recycles information that suggests the fee per month for a PhD level version of OpenAI’s smart software will be a modest $20,000 a month. I think the service one of my team registered costs $20.00 per month. What’s with the 20s? Twenty is a pronic number; that is, it can be slapped on a high school math test so students can say it is the product of two consecutive integers. In college I knew a person who was a numerologist. I recall that the meaning of 20 was cooperation.
The interesting part of the Slashdot post was the comments. I scanned them and concluded that some of the commenters saw the high-end service killing jobs for high-end programmers and consultants. Yeah, maybe. Somehow I doubt that a code base that struggles with information related to a widely-used messaging application is suddenly going to replicate the information I have obtained from my sources in Eastern Europe seems a bit of stretch. Heck, ChatGPT could barely do English. Russian? Not a change, but who knows. And for $200,000 it is not likely this dinobaby will take what seems like unappetizing bait.
One commenter allegedly named TheGreatEmu said:
I was about to make a similar comment, but the cost still doesn’t add up. I’m at a national lab with generally much higher overheads than most places, and a postdoc runs us $160k/year fully burdened. And of course the AI sure as h#ll can’t connect cables, turn knobs, solder, titrate, use a drill press, clean, chat with the machinist who doesn’t use email, sneaker net data out of the air-gapped lab, or understand napkin drawings over beer where all real science gets done. Or do anything useful with information that isn’t already present in the training data, and if you’re not pushing past existing knowledge boundaries, you’re not really doing science are you?
My hunch is that this is a PR or marketing play. Let’s face it. With Microsoft cutting off data center builds and Google floundering with cheese, the smart software revolution is muddling forward. The wins are targeted applications in quite specific domains. Yes, gentle reader, that’s why people pay for Chemical Abstracts online. The information is not on the public Internet. The American Chemical Society has information that the super capable AI outfits have not figured as something the non-computational, organic, or inorganic chemist will use from a somewhat volatile outfit. Get something wrong in a nuclear lab and smart software won’t be too helpful if it hallucinates.
Net net: Is everything marketing? At age 80, my answer is, “Absolutely.” Sam AI-Thinks in terms of trillions. Is $20 trillion the next pricing level?
Stephen E Arnold, March 10, 2025
Patents, AI, and Lawyers: Litigators, Start Your Engines
March 7, 2025
Patents can be a useful source of insights, a fact startup Patlytics is banking on. TechCrunch reports, "Patlytics Raises $14M for its Patent Analytics Platform." The firm turbo-charges intellectual property research with bespoke AI. We learn:
"Patlytics’ large language models (LLMs) and generative AI-powered engine are custom-built for IP-related research and other work such as patent application drafting, invention disclosures, invalidity analysis, infringement detection/analysis, Standard Essential Patents (SEPs) analysis, and IP assets portfolio management."
Apparently, the young firm is already meeting with success. We learn:
"The 1-year-old startup said it has seen a 20x increase in ARR and an 18x expansion in its customer base within six months, with a sustained 300% month-over-month growth rate. Patlytics did not disclose how many customers it has but said approximately 50% of its customer base are law firms, and the other half are corporate clients from industries like semiconductors, bio, pharmaceuticals, and more. Additionally, the company now serves customers in South Korea and Japan, and recently launched its first pilot product in London and Germany. Its clients include Abnormal Security, Google, Koch Disruptive Technologies, Quinn Emanuel Urquhart & Sullivan, Richardson Oliver, Reichman Jorgensen Lehman & Feldberg, Xerox, and Young Basile."
That is quite a client roster in such a short time. This round, combined with April’s seed round, brings the companies funding total to $21 million. The firm will put the funds to use hiring new engineers and expanding its products. Based in New York, Patlytics was launched in January, 2024.
Will AI increase patent litigation? Do Tesla Cybertrucks attract attention?
Cynthia Murrell, March 7, 2025
Another New Search System with AI Too
March 7, 2025
There’s a new AI engine in town down specifically designed to assist with research. The Next Web details the newest invention that comes from a big name in the technology industry: “Tech mogul Launches AI Research Engine Corpora.ai.” Mel Morris is a British tech mogul and the man behind the latest research engine: Corpora.ai.
Morris had Corpora.ai designed to provided in-depth research from single prompts. It is also an incredibly fast engine. It can process two million documents per second. Corpora.ai works by reading a prompt then the AI algorithm scans information, including legal documents, news articles, academic papers, and other Web data. The information is then compiled into summaries or reports.
Morris insists that Corpora.ai is a research engine, not a search engine. He invested $15 million of his personal fortune into the project. Morris doesn’t want to compete with other AI projects, instead he wants to form working relationships:
“His funding aims to create a new business model for LLMs. Rather than challenge the leading GenAI firms, Corpora plans to bring a new service to the sector. The research engine can also integrate existing models on the market. ‘We don’t compete with OpenAI, Google, or Deepseek,’ Morris said. ‘The nice thing is, we can play with all of these AI vendors quite nicely. As they improve their models, our output gets better. It’s a really great symbiotic relationship.’
Mel Morris is a self-made businessman who is the former head of King, the Candy Crush game creator. He also owned and sold the dating Web site, uDate. He might see a return on his Corpora.ai investment .
Whitney Grace, March 7, 2025
Attention, New MBAs in Finance: AI-gony Arrives
March 6, 2025
Another post from the dinobaby. Alas, no smart software used for this essay.
I did a couple of small jobs for a big Wall Street outfit years ago. I went to meetings, listened, and observed. To be frank, I did not do much work. There were three or four young, recent graduates of fancy schools. These individuals were similar to the colleagues I had at the big time consulting firm at which I worked earlier in my career.
Everyone was eager and very concerned that their Excel fevers were in full bloom: Bright eyes, earnest expressions, and a gentle but persistent panting in these meetings. Wall Street and Wall Street like firms in London, England, and Los Angeles, California, were quite similar. These churn outfits and deal makers shared DNA or some type of quantum entanglement.
These “analysts” or “associates” gathered data, pumped it into Excel spreadsheets set up by colleagues or technical specialists. Macros processed the data and spit out tables, charts, and graphs. These were written up as memos, reports for those with big sticks, or senior deciders.
My point is that the “work” was done by cannon fodder from well-known universities business or finance programs.
Well, bad news, future BMW buyers, an outfit called PublicView.ai may have curtailed your dreams of a six figure bonus in January or whatever month is the big momma at your firm. You can take a look at example outputs and sign up free at https://www.publicview.ai/.
If the smart product works as advertised, a category of financial work is going to be reshaped. It is possible that fewer analyst jobs will become available as the gathering and importing are converted to automated workflows. The meetings and the panting will become fewer and father between.
I don’t have data about how many worker bees power the Wall Street type outfits. I showed up, delivered information when queried, departed, and sent a bill for my time and travel. The financial hive and its quietly buzzing drones plugged away 10 or more hours a day, mostly six days a week.
The PublicView.ai FAQ page answers some basic questions; for example, “Can I perform quantitative analysis on the files?” The answer is:
Yes, you can ask Publicview to perform computations on the files using Python code. It can create graphs, charts, tables and more.
This is good news for the newly minted MBAs with programming skills. The bad news is that repeatable questions can be converted to workflows.
Let’s assume this product is good enough. There will be no overnight change in the work for existing employees. But slowly the senior managers will get the bright idea of hiring MBAs with different skills, possibly on a contract basis. Then the work will begin to shift to software. At some point in the not-to-distant future, jobs for humans will be eliminated.
The question is, “How quickly can new hires make themselves into higher value employees in what are the early days of smart software?”
I suggest getting on a fast horse and galloping forward. Donkeys with Excel will fall behind. Software does not require health care, ever increasing inducements, and vacations. What’s interesting is that at some point many “analyst” jobs, not just in finance, will be handled by “good enough” smart software.
Remember a 51 percent win rate from code that does not hang out with a latte will strike some in carpetland as a no brainer. The good news is that MBAs don’t have a graduate degree in 18th century buttons or the Brutalist movement in architecture.
Stephen E Arnold, March 6, 2025
Lawyers and High School Students Cut Corners
March 6, 2025
Cost-cutting lawyers beware: using AI in your practice may make it tough to buy a new BMW this quarter. TechSpot reports, "Lawyer Faces $15,000 Fine for Using Fake AI-Generated Cases in Court Filing." Writer Rob Thubron tells us:
"When representing HooserVac LLC in a lawsuit over its retirement fund in October 2024, Indiana attorney Rafael Ramirez included case citations in three separate briefs. The court could not locate these cases as they had been fabricated by ChatGPT."
Yes, ChatGPT completely invented precedents to support Ramirez’ case. Unsurprisingly, the court took issue with this:
"In December, US Magistrate Judge for the Southern District of Indiana Mark J. Dinsmore ordered Ramirez to appear in court and show cause as to why he shouldn’t be sanctioned for the errors. ‘Transposing numbers in a citation, getting the date wrong, or misspelling a party’s name is an error,’ the judge wrote. ‘Citing to a case that simply does not exist is something else altogether. Mr Ramirez offers no hint of an explanation for how a case citation made up out of whole cloth ended up in his brief. The most obvious explanation is that Mr Ramirez used an AI-generative tool to aid in drafting his brief and failed to check the citations therein before filing it.’ Ramirez admitted that he used generative AI, but insisted he did not realize the cases weren’t real as he was unaware that AI could generate fictitious cases and citations."
Unaware? Perhaps he had not heard about the similar case in 2023. Then again, maybe he had. Ramirez told the court he had tried to verify the cases were real—by asking ChatGPT itself (which replied in the affirmative). But that query falls woefully short of the due diligence required by the Federal Rule of Civil Procedure 11, Thubron notes. As the judge who ultimately did sanction the firm observed, Ramirez would have noticed the cases were fiction had his attempt to verify them ventured beyond the ChatGPT UI.
For his negligence, Ramirez may face disciplinary action beyond the $15,000 in fines. We are told he continues to use AI tools, but has taken courses on its responsible use in the practice of law. Perhaps he should have done that before building a case on a chatbot’s hallucinations.
Cynthia Murrell, March 6, 2025
Shocker! Students Use AI and Engage in Sex, Drugs, and Rock and Roll
March 5, 2025
The work of a real, live dinobaby. Sorry, no smart software involved. Whuff, whuff. That’s the sound of my swishing dino tail. Whuff.
I read “Surge in UK University Students Using AI to Complete Work.” The write up says:
The number of UK undergraduate students using artificial intelligence to help them complete their studies has surged over the past 12 months, raising questions about how universities assess their work. More than nine out of 10 students are now using AI in some form, compared with two-thirds a year ago…
I understand the need to create “real” news; however, the information did not surprise me. But the weird orange newspaper tosses in this observation:
Experts warned that the sheer speed of take-up of AI among undergraduates required universities to rapidly develop policies to give students clarity on acceptable uses of the technology.
As a purely practical matter, information has crossed my about professors cranking out papers for peer review or the ever-popular gray literature consumers that are not reproducible, contain data which have been shaped like a kindergartener’s clay animal, and links to pals who engage in citation boosting.
Plus, students who use Microsoft have a tough time escaping the often inept outputs of the Redmond crowd. A Google user is no longer certain what information is created by a semi reputable human or a cheese-crazed Google system. Emails write themselves. Message systems suggest emojis. Agentic AIs take care of mum’s and pop’s questions about life at the uni.
The topper for me was the inclusion in the cited article of this statement:
it was almost unheard of to see such rapid changes in student behavior…
Did this fellow miss drinking, drugs, staying up late, and sex on campus? How fast did those innovations take to sweep through the student body?
I liked the note of optimism at the end of the write up. Check this:
Janice Kay, a director of a higher education consulting firm: ““There is little evidence here that AI tools are being misused to cheat and play the system. [But] there are quite a lot of signs that will pose serious challenges for learners, teachers and institutions and these will need to be addressed as higher education transforms,” she added.”
That encouraging. The academic research crowd does one thing, and I am to assume that students will do everything the old-fashioned way. When you figure out how to remove smart software from online systems and local installations of smart helpers, let me know. Fix up AI usage and then turn one’s attention to changing student behavior in the drinking, sex, and drug departments too.
Good luck.
Stephen E Arnold, March 5, 2025
Mathematics Is Going to Be Quite Effective, Citizen
March 5, 2025
This blog post is the work of a real-live dinobaby. No smart software involved.
The future of AI is becoming more clear: Get enough people doing something, gather data, and predict what humans will do. What if an individual does not want to go with the behavior of the aggregate? The answer is obvious, “Too bad.”
How do I know that as a handful of organizations will use their AI in this manner? I read “Spanish Running of the Bulls’ Festival Reveals Crowd Movements Can Be Predictable, Above a Certain Density.” If the data in the report are close to the pin, AI will be used to predict and then those predictions can be shaped by weaponized information flows. I got a glimpse of how this number stuff works when I worked at Halliburton Nuclear with Dr. Jim Terwilliger. He and a fellow named Julian Steyn were only too happy to explain that the mathematics used for figuring out certain nuclear processes would work for other applications as well. I won’t bore you with comments about the Monte Carl method or the even older Bayesian statistics procedures. But if it made certain nuclear functions manageable, the approach was okay mostly.
Let’s look at what the Phys.org write up says about bovines:
Denis Bartolo and colleagues tracked the crowds of an estimated 5,000 people over four instances of the San Fermín festival in Pamplona, Spain, using cameras placed in two observation spots in the plaza, which is 50 meters long and 20 meters wide. Through their footage and a mathematical model—where people are so packed that crowds can be treated as a continuum, like a fluid—the authors found that the density of the crowds changed from two people per square meter in the hour before the festival began to six people per square meter during the event. They also found that the crowds could reach a maximum density of 9 people per square meter. When this upper threshold density was met, the authors observed pockets of several hundred people spontaneously behaving like one fluid that oscillated in a predictable time interval of 18 seconds with no external stimuli (such as pushing).
I think that’s an important point. But here’s the comment that presages how AI data will be used to control human behavior. Remember. This is emergent behavior similar to the hoo-hah cranked out by the Santa Fe Institute crowd:
The authors note that these findings could offer insights into how to anticipate the behavior of large crowds in confined spaces.
Once probabilities allow one to “anticipate”, it follows that flows of information can be used to take or cause action. Personally I am going to make a note in my calendar and check in one year to see how my observation turns out. In the meantime, I will try to keep an eye on the Sundars, Zucks, and their ilk for signals about their actions and their intent, which is definitely concerned with individuals like me. Right?
Stephen E Arnold, March 5, 2025
We Have to Spread More Google Cheese
March 4, 2025
A Super Bowl ad is a big deal for companies that shell out for those pricy spots. So it is a big embarrassment when one goes awry. The BBC reports, “Google Remakes Super Bowl Ad After AI Cheese Gaffe.” Google was trying to how smart Gemini is. Instead, the ad went out with a stupid mistake. Writers Graham Fraser and Tom Singleton tell us:
“The commercial – which was supposed to showcase Gemini’s abilities – was created to be broadcast during the Super Bowl. It showed the tool helping a cheesemonger in Wisconsin write a product description by informing him Gouda accounts for ’50 to 60 percent of global cheese consumption.’ However, a blogger pointed out on X that the stat was ‘unequivocally false’ as the Dutch cheese was nowhere near that popular.”
In fact, cheddar and mozzarella vie for the world’s favorite cheese. Gouda is not even a contender. Though the company did remake the ad, one top Googler at first defended Gemini with some dubious logic. We learn:
“Replying to him, Google executive Jerry Dischler insisted this was not a ‘hallucination’ – where AI systems invent untrue information – blaming the websites Gemini had scraped the information from instead. ‘Gemini is grounded in the Web – and users can always check the results and references,’ he wrote. ‘In this case, multiple sites across the web include the 50-60% stat.'”
Sure, users can double check an AI’s work. But apparently not even Google itself can be bothered. Was the company so overconfident it did not use a human copyeditor? Or do those not exist anymore? Wrong information is wrong information, whether technically a hallucination or not. Spitting out data from unreliable sources is just as bad as making stuff up. Google still has not perfected the wildly imperfect Gemini, it seems.
Cynthia Murrell, February 28, 2025
Big Thoughts On How AI Will Affect The Job Market
March 4, 2025
Every time there is an advancement in technology, humans are fearful they won’t make an income. While some jobs disappeared, others emerged and humans adapted to the changes. We’ll continue to adapt as AI becomes more integral in society. How will we handle the changes?
Anthropic, a big player in the OpenAI field, launched The Anthropic Index to understand AI’s effects on labor markers and the economy. Anthropic claims it’s gathering “first-of-its” kind data from Claude.ai anonymized conversations. This data demonstrates how AI is incorporated into the economy. The organization is also building an open source dataset for researchers to use and build on their findings. Anthropic surmises that this data will help develop policy on employment and productivity.
Anthropic reported on their findings in their first paper:
• “Today, usage is concentrated in software development and technical writing tasks. Over one-third of occupations (roughly 36%) see AI use in at least a quarter of their associated tasks, while approximately 4% of occupations use it across three-quarters of their associated tasks.
• AI use leans more toward augmentation (57%), where AI collaborates with and enhances human capabilities, compared to automation (43%), where AI directly performs tasks.
• AI use is more prevalent for tasks associated with mid-to-high wage occupations like computer programmers and data scientists, but is lower for both the lowest- and highest-paid roles. This likely reflects both the limits of current AI capabilities, as well as practical barriers to using the technology.”
The Register put the Anthropic report in layman’s terms in the article, “Only 4 Percent Of Jobs Rely Heavily On AI, With Peak Use In Mid-Wage Roles.” They share that only 4% of jobs rely heavily on AI for their work. These jobs use AI for 75% of their tasks. Overall only 36% of jobs use AI for 25% of their tasks. Most of these jobs are in software engineering, media industries, and educational/library fields. Physical jobs use AI less. Anthropic also found that 57% of these jobs use AI to augment human tasks and 43% automates them.
These numbers make sense based on AI’s advancements and limitations. It’s also common sense that mid-tier wage roles will be affected and not physical or highly skilled labor. The top tier will surf on money; the water molecules are not so lucky.
Whitney Grace, March 4, 2025
AI Summaries Get News Wrong
February 28, 2025
With big news stories emerging at a frantic pace, one might turn to AI to consolidate the key points. If so, one might become woefully ill informed. “AI Chatbots Unable to Accurately Summarise News, BBC Finds.” The BBC tested the biggest AIs on content from its own site–OpenAI’s ChatGPT, Microsoft’s Copilot, Google’s Gemini and Perplexity AI all sat for the exam. None of them passed it, though ChatGPT and Perplexity were less bad than Copilot and Gemini. Tech reporter Imran Rahman-Jones tells us:
“In the study, the BBC asked ChatGPT, Copilot, Gemini and Perplexity to summarise 100 news stories and rated each answer. It got journalists who were relevant experts in the subject of the article to rate the quality of answers from the AI assistants. It found 51% of all AI answers to questions about the news were judged to have significant issues of some form. Additionally, 19% of AI answers which cited BBC content introduced factual errors, such as incorrect factual statements, numbers and dates.”
But it was not just about mixing up, or inventing, facts. The chatbots also struggled with the concept of context and the distinction between facts and opinions. We learn:
“The report said that as well as containing factual inaccuracies, the chatbots ‘struggled to differentiate between opinion and fact, editorialised, and often failed to include essential context’.”
To illustrate the findings, the article gives us a few examples:
- “Gemini incorrectly said the NHS did not recommend vaping as an aid to quit smoking.
- ChatGPT and Copilot said Rishi Sunak and Nicola Sturgeon were still in office even after they had left.
- Perplexity misquoted BBC News in a story about the Middle East, saying Iran initially showed ‘restraint’ and described Israel’s actions as ‘aggressive’.”
So, dear readers, we suggest you take the time to read the news for yourselves. Or, at the very least, get your recaps from another human.
Cynthia Murrell, February 28, 2025
-
- Subscribe to Beyond Search
Feature archive
News archive
-
