The Very Expensive AI Horse Race
December 4, 2024
This write up is from a real and still-alive dinobaby. If there is art, smart software has been involved. Dinobabies have many skills, but Gen Z art is not one of them.
One of the academic nemeses of smart software is a professional named Gary Marcus. Among his many intellectual accomplishments is cameo appearance on a former Jack Benny child star’s podcast. Mr. Marcus contributes his views of smart software to the person who, for a number of years, has been a voice actor on the Simpsons cartoon.
The big four robot stallions are racing to a finish line. Is the finish line moving away from the equines faster than the steeds can run? Thanks, MidJourney. Good enough.
I want to pay attention to Mr. Marcus’ Substack post “A New AI Scaling Law Shell Game?” The main idea is that the scaling law has entered popular computer jargon. Once the lingo of Galileo, scaling law now means that AI, like CPUs, are part of the belief that technology just gets better as it gets bigger.
In this essay, Mr. Marcus asserts that getting bigger may not work unless humanoids (presumably assisted by AI0 innovate other enabling processes. Mr. Marcus is aware of the cost of infrastructure, the cost of electricity, and the probable costs of exhausting content.
From my point of view, a bit more empirical “evidence” would be useful. (I am aware of academic research fraud.) Also, Mr. Marcus references me when he says keep your hands on your wallet. I am not sure that a fix is possible. The analogy is the old chestnut about changing a Sopwith Camel’s propeller when the aircraft is in a dogfight and the synchronized machine gun is firing through the propeller.
I want to highlight one passage in Mr. Marcus’ essay and offer a handful of comments. Here’s the passage I noted:
Over the last few weeks, much of the field has been quietly acknowledging that recent (not yet public) large-scale models aren’t as powerful as the putative laws were predicting. The new version is that there is not one scaling law, but three: scaling with how long you train a model (which isn’t really holding anymore), scaling with how long you post-train a model, and scaling with how long you let a given model wrestle with a given problem (or what Satya Nadella called scaling with “inference time compute”).
I think this is a paragraph I will add to my quotes file. The reasons are:
First, investors, would be entrepreneurs, and giant outfits really want a next big thing. Microsoft fired the opening shot in the smart software war in early 2023. Mr. Nadella suggested that smart software would be the next big thing for Microsoft. The company has invested in making good on this statement. Now Microsoft 365 is infused with smart software and Azure is burbling with digital glee with its “we’re first” status. However, a number of people have asked, “Where’s the financial payoff?” The answer is standard Silicon Valley catechism: The payoff is going to be huge. Invest now.” If prayers could power hope, AI is going to be hyperbolic just like the marketing collateral for AI promises. But it is almost 2025, and those billions have not generated more billions and profit for the Big Dogs of AI. Just sayin’.
Second, the idea that the scaling law is really multiple scaling laws is interesting. But if one scaling law fails to deliver, what happens to the other scaling laws? The interdependencies of the processes for the scaling laws might evoke new, hitherto identified scaling laws. Will each scaling law require massive investments to deliver? Is it feasible to pay off the investments in these processes with the original concept of the scaling law as applied to AI. I wonder if a reverse Ponzi scheme is emerging. The more pumped in the smaller the likelihood of success. Is AI a demonstration of convergence or The mathematical property you’re describing involves creating a sequence of fractions where the numerator is 1 and the denominator is an increasing sequence of integers. Just askin’.
Third, the performance or knowledge payoff I have experienced with my tests of OpenAI and the software available to me on You.com makes clear that the systems cannot handle what I consider routine questions. A recent example was my request to receive a list of the exhibitors at the November 1 Gateway Conference held in Dubai for crypto fans of Telegram’s The Open Network Foundation and TON Social. The systems were unable to deliver the lists. This is just one notable failure which a humanoid on my research team was able to rectify in an expeditious manner. (Did you know the Ku Group was on my researcher’s list?) Just reportin’.
Net net: Will AI repay the billions sunk into the data centers, the legal fees (many still looming), the staff, and the marketing? If you ask an accelerationist, the answer is, “Absolutely.” If you ask a dinobaby, you may hear, “Maybe, but some fundamental innovations are going to be needed.” If you ask an AI will kill us all type like the Xoogler Mo Gawdat, you will hear, “Doom looms.” Just dinobabyin’.
Stephen E Arnold, December 4, 2024
Legacy Code: Avoid, Fix, or Flee (Two Out of Three Mean Forget It)
December 4, 2024
In his Substack post, “Legacy Schmegacy,” software engineer David Reis offers some pointers on preventing and coping with legacy code. We found this snippet interesting:
“Someone must fix the legacy code, but it doesn’t have to be you. It’s far more honorable to switch projects or companies than to lead a misguided rewrite.”
That’s the spirit: quit and let someone else deal with it. But not everyone is in the position to cut and run. For those actually interested in addressing the problem, Reis has some suggestions. First, though, the post lists factors that can prevent legacy code in the first place:
- “The longer a programmer’s tenure the less code will become legacy, since authors will be around to appreciate and maintain it.
- The more code is well architected, clear and documented the less it will become legacy, since there is a higher chance the author can transfer it to a new owner successfully.
- The more the company uses pair programming, code reviews, and other knowledge transfer techniques, the less code will become legacy, as people other than the author will have knowledge about it.
- The more the company grows junior engineers the less code will become legacy, since the best way to grow juniors is to hand them ownership of components.
- The more a company uses simple standard technologies, the less likely code will become legacy, since knowledge about them will be widespread in the organization. Ironically if you define innovation as adopting new technologies, the more a team innovates the more legacy it will have. Every time it adopts a new technology, either it won’t work, and the attempt will become legacy, or it will succeed, and the old systems will.”
Reiss’ number one suggestion to avoid creating legacy code is, “don’t write crappy code.” Noted. Also, stick with tried and true methods unless shiny a new tech is definitely the best option. Perhaps most importantly, coders should teach others in the organization how their code works and leave behind good documentation. So, common sense and best practices. Brilliant!
When confronted with a predecessor’s code, he advises one to “delegacify” it. That is a word he coined to mean: Take time to understand the code and see if it can be improved over time before tossing it out entirely. Or, as noted above, just run away. That can be an option for some.
Cynthia Murrell, December 4, 2024
The New Coca Cola of Marketing: Xmas Ads
December 4, 2024
Though Coca-Cola has long purported that “It’s the Real Thing,” a recent ad is all fake. NBC News reports, “Coca-Cola Causes Controversy with AI-Made Ad.” We learn:
“Coca-Cola is facing backlash online over an artificial intelligence-made Christmas promotional video that users are calling ‘soulless’ and ‘devoid of any actual creativity.’ The AI-made video features everything from big red Coca-Cola trucks driving through snowy streets to people smiling in scarves and knitted hats holding Coca-Cola bottles. The video was meant to pay homage to the company’s 1995 commercial ‘Holidays Are Coming,’ which featured similar imagery, but with human actors and real trucks.”
The company’s last ad generated with AI, released earlier this year, did not face similar backlash. Is that because, as University of Wisconsin-Madison’s Neeraj Arora suggests, Coke’s Christmas ads are somehow sacrosanct? Or is it because March’s Masterpiece is actually original, clever, and well executed? Or because the artworks copied in that ad are treated with respect and, for some, clearly labeled? Whatever the reason, the riff on Coca-Cola’s own classic 1995 ad missed the mark.
Perhaps it was just too soon. It may be a matter of when, not if, the public comes to accept AI-generated advertising as the norm. One thing is certain: Coca Cola knows how to make sure marketing professors teach memorable case examples of corporate “let’s get hip” thinking.
Cynthia Murrell, December 4, 2024
FOGINT: Telegram and Its Race Against Time
December 3, 2024
The article is a product of the humans working on the FOGINT team. The image is from Gifr.com.
The Financial Times recently stirred debate in the cryptocurrency community with its article, “Telegram Finances Propped Up by Crypto Gains As Founder Fights Charges.” Telegram’s ambitions for an initial public offering (IPO) hinge on proving it has a sustainable, profitable business model.
According to the FT, Telegram sold off cryptocurrency holdings to shore up its balance sheet, reporting revenue of $525 million. The financials, based on unaudited statements, framed the crypto sale as a “tactical” move, with Durov’s confinement in France having no material impact on the company.
CCN added flair with its piece, “Pavel Durov’s Telegram Nets $335M Windfall: Can It Ride a Crypto Bull to a $30B IPO by 2026,” highlighting that while crypto revenues and TON reserves have helped Telegram stay afloat, the firm still faces substantial debt and operating losses—$259 million in 2023 alone.
Crypto.news zeroed in on debt in its article, “How Telegram Made Over Half a Billion Dollars Thanks to Crypto?” Telegram, wholly owned by Durov, has raised $2.4 billion in debt financing, with repayment looming in 2026. In September, it used part of its crypto proceeds to repurchase $124.5 million in bonds. (Note: None of the news sources we reviewed noted that Telegram is using a variant of the MicroStrategy strategy of acquiring crypto currency to pump up the company’s “value.” See DLNews for more detail.)
The FOGINT research team identifies three key dynamics:
- Cash Flow Through Crypto Sales: Telegram’s crypto transactions inject much-needed liquidity, transforming a significant 2023 loss into a manageable red blot on its financial history.
- Tight Links to The Open Network Foundation (TON): Despite TON’s ostensible independence, the Foundation is deeply intertwined with Telegram. This relationship traces back to the U.S. SEC’s 2019 intervention against GRAM (now TON), which forced Telegram to offload its blockchain to the “independent” Foundation in 2023. Regulators in the U.S., UAE, and Switzerland appear to tolerate this arrangement for now.
- Racing Against the Clock: Telegram is fast-tracking innovations, like the BotFather’s high-speed processing and partnerships with firms such as Ku Group. Its developer meetups and funding programs are designed to rapidly build out its ecosystem. The urgency stems from a softening stance toward law enforcement—Telegram now appears more willing to share user data, potentially feeding into global investigative pipelines.
This newfound openness aligns with Telegram’s aggressive push to monetize its TON blockchain. At the November 1-2 Gateway Conference in Dubai, Telegram launched an all-out campaign to promote its crypto ecosystem. From YouTube videos and meetups to venture fund pitches, the effort signals a company operating in overdrive.
Blockchain researcher Sean Brizendine said: “”The Telegram hype is definitely real, and the Durov brothers’ future is on the line. Now is the time to pay attention as Telegram’s moves are breaking fast and furious.”
Why the rush? Telegram’s 900 million users and its wildly popular crypto games have been critical growth engines. But with its ethos of “do what you want” giving way to “use our crypto platform,” the stakes are higher than ever. The “leader” is in the grasp of French authorities. Fast-fashion-like life cycles of its crypto ventures underline a harsh reality: For Telegram, succeeding in crypto is no longer optional—it’s a turning point for the company and its affiliated organizations. Cornered animals can be more dangerous than some people think.
Stephen E Arnold, December 3, 2024
BlueSky: Tweeting Birds Are Flocking Around
December 3, 2024
As X, formerly Twitter, becomes more toxic, alternative BlueSky has welcomed many refugees fleeing Musk’s regime. In fact, the decentralized social media platform recently hit 15 million users. Blood in the Machine takes this opportunity to declare, “Bluesky’s Success Is a Rejection of Big Tech’s Operating System.” The post is largely about enumerating X’s flaws. Will such a marketing angle make the Twitter clone a winner?
After a brief lesson in recent social-media history, blogger Brian Merchant observes:
“The online world has become so hostile to users that Bluesky’s pitch of ‘here is a straightforward feed of text-based user-generated posts that we promise not to mess with’ is revelatory. Its scaling model and raison d’être are a very rejection of the platforms that have colonized the rest of our digital lives, and relentlessly commodified them. No wonder everyone seems to be rooting for its success, even if there are, pointedly, no guarantees those ideals will remain in place.”
Why no guarantees? The taint of venture capital, for one. The platform’s recent $15 million series A funding round was led by Blockchain Capital. Despite that firm’s focus on cryptocurrency, BlueSky promises it will continue to prioritize the user over the likes of crypto and NFTs. Will it deliver? At least wary users can turn to Mastodon. For now.
The write-up continues:
“BlueSky is giving hope to people who spend long hours online precisely because it is purporting to be, and so far succeeding, at least in its very short lifespan, in being everything that big tech is not. No AI spam, no glitchy ad tech, no link throttling, no malignant billionaire owner. BlueSky is not just tapping into this wellspring of goodwill because it promises a return to the halcyon days of *Twitter*—but a return to the days before ossified, rent-seeking tech monopolies drove our collective online experience to hell.”
But how long will this retro trip last?
Cynthia Murrell, December 3, 2024
New Concept: AI High
December 3, 2024
Is the AI hype-a-thon finally slowing? Nope. And our last nerves may not be the only thing to suffer. The AI industry could be shooting itself in the foot. ComputerWorld predicts, “AI Is on a Fast Track, but Hype and Immaturity Could Derail It.” Writer Scot Finnie reports:
“The hype is so extreme that a fall-out, which Gartner describes in its technology hype cycle reports as the ‘trough of disillusionment,’ seems inevitable and might be coming this year. That’s a testament to both genAI’s burgeoning potential and a sign of the technology’s immaturity. The outlook for deep learning for predictive models and genAI for communication and content generation is bright. But what’s been rarely mentioned amid the marketing blitz of recent months is that the challenges are also formidable. Machine learning tools are only as good as the data they’re trained with. Companies are finding that the millions of dollars they’ve spent on genAI have yielded lackluster ROI because their data is filled with contradictions, inaccuracies, and omissions. Plus, the hype surrounding the technology makes it difficult to see that many of the claimed benefits reside in the future, not the present.”
Oops. The article notes some of the persistent problems with generative AI, like hallucinations, repeated crashes ,and bias. Then there are the uses bad actors have for these tools, from phishing scams to deepfakes. For investors, disappointing results and returns are prompting second thoughts. None of this means AI is completely worthless, Finnie advises. He just suggests holding off until the rough edges are smoothed out before going all in. Probably a good idea. Digital mushrooms.
December 3, 2024
Pass a Law to Prevent Youngsters from Accessing Social Media. Yep, That Will Work Well
December 2, 2024
This is the work of a dinobaby. Smart software helps me with art, but the actual writing? Just me and my keyboard.
I spotted a very British “real” news story called “It’s So Easy to Lie: : A Fifth of Children Use Fake Age on Social Media.” I like the idea that one can pick 100 children at random from a school with 13 year olds, only 80 percent will allegedly follow the rules.
Thanks, Midjourney. Good enough. I might point out you did not present a young George Washington despite my efforts to feed you words to which you would respond.
Does the 20 percent figure seem low to you? I would suggest that if a TikTok-type video was popular at that school, more than 20 percent would find a way to get access to that video. If the video was about being thin or a fashion tip, the females would be more interested and they would lie to get that information. The boys might be more interested in other topics, which I shall leave to your imagination.
The write up says:
A newly released survey, conducted by the UK media regulator, indicates 22% of eight to 17 year olds lie that they are 18 or over on social media apps.
I doubt that my hypothetical group of 13 years olds are different from those who are four years older. The write up pointed out:
A number of tech firms have recently announced measures to make social media safer for young people, such as Instagram launching “teen accounts.” However, when BBC news spoke to a group of teenagers at Rosshall Academy, in Glasgow, all of them said they used adult ages for their social media accounts. “It’s just so easy to lie about your age”, said Myley, 15.
Australia believes it has a fix: Ban access. I quite like the $AUS 33 million fine too.
I would suggest that in a group of 100 teens, one will know how to create a fake persona, buy a fake ID from a Telegram vendor, and get an account. Will a Telegram user set up a small online business to sell fake identities or social media accounts to young people? Yep.
Cyber security firms cannot block bad actors. What makes regulators think that social media companies can prevent young people from getting access to their service. Enjoy those meetings. I hope the lunches are good.
My hunch is that the UK is probably going to ban social media access for those under a certain age. Good luck.
Stephen E Arnold, December 2, 2024
FOGINT: Telegram and Its Possible Latent Weakness
December 2, 2024
This write up is the work of a dinobaby. Thanks to Gifer.com for the moving fog!
Cointelegram gathered some interesting information about Telegram. Let’s take a look at some of the data in Cointelegraph’s “Telegram’s Crypto Holdings Rose to $1.3B in H1 2024,” November 25 or 26, 2024. Verification of the data is difficult, which is a frequent issue where crypto valuations are presented. Let’s assume that the general information reflects Telegram’s business position.,
According to the article:
Telegram’s crypto holdings jumped from $400 million to $1.3 billion in H1 2024, driven by Toncoin sales and strategic deals.
Translating: Telegram dumped some crypto, made a profit, and reported the upside to fatten its financial position.
The key point is that Telegram allegedly had about $400 million in digital assets. Now the company has more than $1 billion.
Allegedly Telegram generated about $500 million in revenue in the period from January to June 2024. The boost does not come from advertising and subscriptions. Crypto dealing and other business agreements have bolstered the company’s apparently positive financials. It is worth noting that an audit of Telegram’s finances is rumored to have reported that Telegram lost money in a previous financial period.
The Cointelegraph story reminded its readers that Pavel Durov is confined to France. His legal issues have not been resolved. He has posted bail in the neighborhood of $5 million euros, an indication of the seriousness of the French charges against him.
The FOGINT’s research team notes:
- Telegram is making significant moves with its tie ups with organizations like CryptoCasino.com. This is part of the firm’s effort to become the platform for Telegram-centric gaming
- The Open Network Foundation (which runs on the Telegram platform) continues to promote the TON crypto and the Telegram-centric capabilities on which the Foundation’s services run. One of these initiatives involves investing in TON-centric applications and holding training courses in major international cities.
- Telegram’s “value” as an allegedly secure messaging app is eroding. Mr. Durov has insisted that Telegram cooperates with law enforcement. Those statements mean that Telegram has to kick its online gaming activities and the Foundation’s financial services into high gear.
Net net: Telegram may be strong at first glance, but there may be a latent weakness in the Telegram, Foundation, and TON.Social entities.
Stephen E Arnold, December 2, 2024
Deepfakes: An Interesting and Possibly Pernicious Arms Race
December 2, 2024
As it turns out, deepfakes are a difficult problem to contain. Who knew? As victims from celebrities to schoolchildren multiply exponentially, USA Today asks, “Can Legislation Combat the Surge of Non-Consensual Deepfake Porn?” Journalist Dana Taylor interviewed UCLA’s John Villasenor on the subject. To us, the answer is simple: Absolutely not. As with any technology, regulation is reactive while bad actors are proactive. Villasenor seems to agree. He states:
“It’s sort of an arms race, and the defense is always sort of a few steps behind the offense, right? In other words that you make a detection tool that, let’s say, is good at detecting today’s deepfakes, but then tomorrow somebody has a new deepfake creation technology that is even better and it can fool the current detection technology. And so then you update your detection technology so it can detect the new deepfake technology, but then the deepfake technology evolves again.”
Exactly. So if governments are powerless to stop this horror, what can? Perhaps big firms will fight tech with tech. The professor dreams:
“So I think the longer term solution would have to be automated technologies that are used and hopefully run by the people who run the servers where these are hosted. Because I think any reputable, for example, social media company would not want this kind of content on their own site. So they have it within their control to develop technologies that can detect and automatically filter some of this stuff out. And I think that would go a long way towards mitigating it.”
Sure. But what can be done while we wait on big tech to solve the problem it unleased? Individual responsibility, baby:
“I certainly think it’s good for everybody, and particularly young people these days to be just really aware of knowing how to use the internet responsibly and being careful about the kinds of images that they share on the internet. … Even images that are sort of maybe not crossing the line into being sort of specifically explicit but are close enough to it that it wouldn’t be as hard to modify being aware of that kind of thing as well.”
Great, thanks. Admitting he may sound naive, Villasenor also envisions education to the (partial) rescue:
“There’s some bad actors that are never going to stop being bad actors, but there’s some fraction of people who I think with some education would perhaps be less likely to engage in creating these sorts of… disseminating these sorts of videos.”
Our view is that digital tools allow the dark side of individuals to emerge and expand.
Cynthia Murrell, December 2, 2024
The Golden Fleecer of the Year: Boeing
November 29, 2024
When I was working in Washington, DC, I had the opportunity to be an “advisor” to the head of the Joint Committee on Atomic Energy. I recall a comment by Craig Hosmer (R. California) and retired rear admiral saying, “Those Air Force guys overpay.” The admiral was correct, but I think that other branches of the US Department of Defense have been snookered a time or two.
In the 1970s and 1980s, Senator William Proxmire (D. Wisconsin) had one of his staff keep an eye of reports about wild and crazy government expenditures. Every year, the Senator reminded people of a chivalric award dating allegedly from the 1400s. Yep, the Middle Ages in DC.
The Order of the Golden Fleece in old timey days of yore meant the recipient received a snazzy chivalric order intended to promote Christian values and the good neighbor policy of Spain and Austria. A person with the fleece was important, a bit like a celebrity arriving at a Hollywood Oscar event. (Yawn)
Thanks, Wikipedia. Allegedly an example of a chivalric Golden Fleece. Yes, that is a sheep, possibly dead or getting ready to be dipped. Thanks,
Reuters, the trusted outfit which tells me it is trusted each time I read one of its “real” news stories, published “Boeing Overcharged Air Force Nearly 8,000% for Soap Dispensers, Watchdog Alleges.” The write up stated in late October 2024:
Boeing overcharged the U.S. Air Force for spare parts for C-17 transport planes, including marking up the price on soap dispensers by 7,943%, according to a report by a Pentagon watchdog. The Department of Defense Office of Inspector General said on Tuesday the Air Force overpaid nearly $1 million for a dozen spare parts, including $149,072 for an undisclosed number of lavatory soap dispensers from the U.S. plane maker and defense contractor.
I have heard that the Department of Defense has not been able to monitor some of its administrative activities or complete an audit of what it does with its allocated funds.
According to the trusted write up:
The Pentagon’s budget is huge, breaking $900 billion last year, making overcharges by defense contractors a regular headache for internal watchdogs, but one that is difficult to detect. The Inspector General also noted it could not determine if the Air Force paid a fair price on $22 million of spare parts because the service did not keep a database of historical prices, obtain supplier quotes or identify commercially similar parts.
My view is that one of the elected officials in Washington, DC, should consider reviving the Proxmire Golden Fleece Award. Boeing may qualify, but there may be other contenders for the award as well.
I quite like the idea of scope changes and engineering change orders for some US government projects. But I have to admit that Senator Proxmire’s identification of a $600 hammer sold to the US Department of Defense is not interesting.
That 8,000 percent mark up is pretty nifty. Oh, on Amazon soap dispensers cost between $20 and $100. Should the Reuters’ story have mentioned:
- Procurement reform
- Poor financial controls
- Lack of common sense?
Of course not! The trust outfit does not get mired in silly technicalities. And Boeing? That outfit is doing a bang up job.
Stephen E Arnold, November 29, 2024
-
- Subscribe to Beyond Search
Feature archive
News archive
-
