In Big Data, Bad Data Does Not Matter. Not So Fast, Mr. Slick
April 8, 2024
This essay is the work of a dumb dinobaby. No smart software required.
When I hear “With big data, bad data does not matter. It’s the law of big numbers. Relax,” I chuckle. Most data present challenges. First, figuring out which data are accurate can be a challenge. But the notion of “relax,” does not cheer me. Then one can consider data which have been screwed up by a bad actor, a careless graduate student, a low-rent research outfit, or someone who thinks errors are not possible.
The young vendor is confident that his tomatoes and bananas are top quality. The color of the fruit means nothing. Thanks, MSFT Copilot. Good enough, like the spoiled bananas.
“Data Quality Getting Worse, Report Says” offers some data (which may or may not be on the mark) which remind me to be skeptical of information available today. The Datanami article points out:
According to the company’s [DBT Labs’] State of Analytics Engineering 2024 report released yesterday, poor data quality was the number one concern of the 456 analytics engineers, data engineers, data analysts, and other data professionals who took the survey. The report shows that 57% of survey respondents rated data quality as one of the three most challenging aspects of the data preparation process. That’s a significant increase from the 2022 State of Analytics Engineering report, when 41% indicated poor data quality was one of the top three challenges.
The write up offers several other items of interest; for example:
- Questions about who owns the data
- Integration of fusion of multiple data sources
- Documenting data products; that is, the editorial policy of the producer / collector of the information.
This flashing yellow light about data seems to be getting brighter. The implication of the report is that data quality “appears” to be be heading downhill. The write up quotes Jignesh Patel, computer science professor at Carnegie Mellon University to underscore the issue:
“Data will never be fully clean. You’re always going to need some ETL [extract, transform, and load] portion. The reason that data quality will never be a “solved problem,” is partly because data will always be collected from various sources in various ways, and partly because or data quality lies in the eye of the beholder. You’re always collecting more and more data. If you can find a way to get more data, and no one says no to it, it’s always going to be messy. It’s always going to be dirty.”
But what about the assertion that in big data, bad data will be a minor problem. That assertion may be based on a lack of knowledge about some of the weak spots in data gathering processes. In the last six months, my team and I have encountered these issues:
- The source of the data contained a flaw so that it was impossible to determine what items were candidates for filtering out
- The aggregator had zero controls because it acquired data from another party and did not homework other than hyping a new data set
- Flawed data filled the exception folder with a large percentage of the information that remediation was not possible due to time and cost constraints
- Automated systems are indiscriminate, and few (sometimes no one) pay close attention to inputs.
I agree that data quality is a concern. However, efficiency trumps old-fashioned controls and checks applied via subject matter experts and trained specialists. The fix will be smart software which will be cheaper and more opaque. The assumption that big data will be self healing may not be accurate, but it sounds good.
Stephen E Arnold, April 8, 2024
HP and Autonomy: The Long Tail of Search and Retrieval
April 8, 2024
This essay is the work of a dumb dinobaby. No smart software required.
The US justice system is flawed but when big money is at stake, it quickly works as it’s supposed to do. British tech entrepreneur is responsible for making the tech industry lose a lot of greenbacks and the BBC shares the details: “Mike Lynch: Autonomy Founder’s Fraud Trial Begins In The US.” Lynch, formerly called Britain’s equivalent of Bill Gates, was extradited to the US in 2023 after a British court found him guilty of a civil fraud cause. He is accused of over inflating the value of his former company Autonomy. Autonomy was sold to Hewlett-Packard (HP) in 2011 for $11 billion.
Lynch is facing sixteen charges and a possible twenty-five years in prison if convicted. Reid Weingarten, Lynch’s attorney, stated his client is prepared to take the stand. He also said that Lynch focused on Autonomy’s technology side and left the finances to others. After buying Autonomy, HP valued it at $2.2 billion and claimed Lynch duped them.
Lynch founded Autonomy in 1996 and it became a top 100 public companies in the United Kingdom. Autonomy was known for software that extracted information from unstructured content: video, emails, and phone calls.
HP is not mincing claims in this case:
“US prosecutors in San Francisco say Mr Lynch backdated agreements to mislead about the company’s sales; concealed the firm’s loss-making business reselling hardware; and intimidated or paid off people who raised concerns, among other claims. In court filings, his attorneys have argued that the "real reason for the write-down" was a failure by HP to manage the merger. ‘Then, with its stock price crumbling under the weight of its own mismanagement, circled the wagons to protect its new leaders and wantonly accused’ Mr Lynch of fraud, they wrote.”
London’s High Court convicted Lynch and Autonomy’s former CFO Sushovan Hussain of fraud. Hussain was imprisoned for five years and fined millions of dollars. The pair claimed HP’s case against them was buyer’s remorse and management failings.
Lynch should be held accountable for false claims, pay the fines, and be jailed if declared guilty. If the court does convict him, it will be time for more legal gymnastics.
Whitney Grace, April 8, 2024
Perplexed at Perplexity? It Is Just the Need for Money. Relax.
April 5, 2024
This essay is the work of a dumb dinobaby. No smart software required.
“Gen-AI Search Engine Perplexity Has a Plan to Sell Ads” makes it clear that the dynamic world of wildly over-hyped smart software is somewhat fluid. Pivoting from “No, never” to “Yes, absolutely” might catch some by surprise. But this dinobaby is ready for AI’s morphability. Artificial intelligence means something to the person using the term. There may be zero correlation between the meaning of AI in the mind of any other people. Absent the Vulcan mind meld, people have to adapt. Morphability is important.
The dinobaby analyst is totally confused. First, say one thing. Then, do the opposite. Thanks, MSFT Copilot. Close enough. How’s that AI reorganization going?
I am thinking about AI because Perplexity told Adweek that despite obtaining $73 million in Series B funding, the company will start selling ads. This is no big deal for Google which slips unmarked ads into its short video streams. But Perplexity was not supposed to sell ads. Yeah, well, that’s no longer an operative concept.
The write up says:
Perplexity also links sources in the response while suggesting related questions users might want to ask. These related questions, which account for 40% of Perplexity’s queries, are where the company will start introducing native ads, by letting brands influence these questions,
Sounds rock solid, but I think that the ads will have a bit of morphability; that is, when big bucks are at stake, those ads are going to go many places. With an alleged 10 million monthly active users, some advertisers will want those ads shoved down the throat of anything that looks like a human or bot with buying power.
Advertisers care about “brand safety.” But those selling ads care about selling ads. That’s why exciting ads turn up in quite interesting places.
I have a slight distrust for pivoters. But that’s just an old dinobaby, an easily confused dinobaby at that.
Stephen E Arnold, April 5, 2024
Nah, AI Is for Little People Too. Ho Ho Ho
April 5, 2024
This essay is the work of a dumb dinobaby. No smart software required.
I like the idea that smart software is open. Anyone can download software and fire up that old laptop. Magic just happens. The reality is that smart software is going to involve some big outfits and big bucks when serious applications or use cases are deployed. How do I know this? Well, I read “Microsoft and OpenAI Reportedly Building $100 Billion Secret Supercomputer to Train Advanced AI.” The number $100 billion in not $6 trillion bandied about by Sam AI-Man a few weeks ago. It does, however, make Amazon’s paltry $3 billion look like chump change. And where does that leave the AI start ups, the AI open source champions, and the plain vanilla big-smile venture folks? The answer is, “Ponying up some bucks to get that AI to take flight.”
Thanks, MSFT Copilot. Stick to your policies.
The write up states:
… the dynamic duo are working on a $100 billion — that’s "billion" with a "b," meaning a sum exceeding many countries’ gross domestic products — on a hush-hush supercomputer designed to train powerful new AI.
The write up asks a question some folks with AI sparkling in their eyes cannot answer; to wit:
Needless to say, that’s a mammoth investment. As such, it shines an even brighter spotlight on a looming question for the still-nascent AI industry: how’s the whole thing going to pay for itself?
But I know the answer: With other people’s money and possibly costs distributed across many customers.
Observations are warranted:
- The cost of smart software is likely to be an issue for everyone. I don’t think “free” is the same as forever
- Mistral wants to do smaller language models, but Microsoft has “invested” in that outfit as well. If necessary, some creative end runs around an acquisition may be needed because MSFT may want to take Mistral off the AI chess board
- What’s the cost of the electricity to operate what $100 billion can purchase? How about a nifty thorium reactor?
Net net: Okay, Google, what is your move now that MSFT has again captured the headlines?
Stephen E Arnold, April 5, 2024
Finding Live Music Performances
April 5, 2024
This essay is the work of a dumb dinobaby. No smart software required.
Here is a niche search category some of our readers will appreciate. Lifehacker shares “The Best Ways to Find Live Gigs for Music You Love.” Writer David Nield describes how one can tap into a combination of sources to stay up to date on upcoming music events. He begins:
“More than once I’ve missed out on shows in my neighborhood put on by bands I like, just because I’ve been out of the loop. Whether you don’t want to miss gigs by artists you know, or you’re keen to get out and discover some new music, there are lots of ways to stay in touch with the live shows happening in your area—you need never miss a gig again. Pick the one(s) that work best for you from this list.”
First are websites dedicated to spreading the musical word, like Songkick and Bandsintown. One can sign up for notices or simply browse the site by artist or location. These sites can also use one’s listening data from streaming apps to inform their results. Or one can go straight to the source and follow artists on social media or their own websites (but that can get overwhelming if one enjoys many bands). Several music apps like Spotify and Deezer will notify you of upcoming concerts and events for artists you choose. Finally, YouTube lists tour details and ticket links beneath videos of currently touring bands, highlighting events near you. If, that is, you have chosen to share your location with the Google-owned site.
Cynthia Murrell, April 5, 2024
McKinsey & Co. Emits the Message “You Are No Longer the Best of the Best”
April 4, 2024
This essay is the work of a dumb dinobaby. No smart software required.
I love blue chip consulting firms’ management tactics. I will not mention the private outfits which go public and then go private. Then the firms’ “best of the best” partners decide to split the firm. Wow. Financial fancy dancing or just evidence that “best of the best” is like those plastic bottles killing off marine life?
I read “McKinsey Is so Eager to Trim Staff That It’s Offering Some Employees 9 Months’ Pay to Go and Do Something Else. I immediately asked myself, “What’s some mean?” I am guessing based on my experience that “all” of the RIF’ed staff are not getting the same deal. Well, that’s life in the exciting world of the best and the brightest. Some have to accept that there are blue chippers better and, therefore, able to labor enthusiastically at a company known as the Big Dog in the consulting world.
Thanks MSFT Copilot. (How’s your security today?)
The write up reports as “real” NY news:
McKinsey is attempting to slim the company down in a caring and supporting way by paying its workers to quit.
Hmmm. “Attempting” seems an odd word for a consulting firm focused on results. One slims down or one remains fat and prone to assorted diseases if I understood my medical professional. Is McKinsey signaling that its profit margin is slipping like the trust level for certain social media companies? Or is artificial intelligence the next big profit making thing; therefore, let’s clear out the deadwood and harvest the benefits of smart software unencumbered by less smart humans?
Plus, the formerly “best and brightest” will get help writing their résumés. My goodness, imagine a less good Type A super achiever unable to write a résumé. But just yesterday those professionals were able to advise executives often with decades more experience, craft reports with asterisk dot points, and work seven days a week. These outstanding professionals need help writing their résumés. This strikes me as paternalistic and a way to sidestep legal action for questionable termination.
Plus, the folks given the chance to find their future elsewhere (as long as the formerly employed wizard conforms to McKinsey’s policies about client poaching) can allegedly use their McKinsey email accounts. What might a person who learns he or she is no longer the best of the best might do with a live McKinsey email account? I have a couple of people on my research team who have studied mischief with emails. I assume McKinsey’s leadership knows a lot more than my staff. We don’t pontificate about pharmaceutical surfing; we do give lectures to law enforcement and intelligence professionals. Therefore, my team knows much, much less about the email usage that McKinsey management.
Deloitte, another blue chip outfit, is moving quickly into the AI space. I have heard that it wants to use AI and simultaneously advise its clients about AI. I wonder if Deloitte has considered that smart software might be marginally less expensive than paying some of the “best of the best” to do manual work for clients? I don’t know.
The blue chip outfit at which I worked long ago was a really humane place. Those rumors that an executive drowned a loved one were just rumors. The person was a kind and loving individual with a raised dais in his office. I recall I hard to look up at him when seated in front of his desk. Maybe that’s just an AI type hallucination from a dinobaby. I do remember the nurturing approach he took when pointing at a number and demanding the VP presenting the document, “I want to know where that came from now.” Yes, that blue chip professional was patient and easy going as well.
I noted this passage in the Fortune “real” NY news:
A McKinsey spokesperson told Fortune that its unusual approach to layoffs is all part of the company’s core mission to help people ‘learn and grow into leaders, whether they stay at McKinsey or continue their careers elsewhere.’
I loved the sentence including the “learn and grow into leaders” verbiage. I am imagining a McKinsey HR professional saying, “Remember when we recruited you? We told you that you were among the top one percent of the top one percent. Come on. I know you remember? Oh, you don’t remember my assurances of great pay, travel, wonderful colleagues, tremendous opportunities to learn, and build your interpersonal skills. Well, that’s why you have been fired. But you can use your McKinsey email. Please, leave now. I have billable work to do that you obviously were not able to undertake and complete in a satisfactory manner. Oh, here’s your going away gift. It is a T shirt which says, ‘Loser@mckinsey.com.’”
Stephen E Arnold, April 4, 2024
Yeah, Stability at Stability AI: Will Flame Outs Light Up the Bubble?
April 4, 2024
This essay is the work of a dumb dinobaby. No smart software required.
I read “Inside the $1 Billion Love Affair between Stability AI’s Complicated Founder and Tech Investors Coatue and Lightspeed—And How It Turned Bitter within Months.” Interesting but, from my point of view, not surprising. High school science club members, particularly when preserving some of their teeny bopper ethos into alleged adulthood can be interesting people. And at work, exciting may be a suitable word. The write up’s main idea is that the wizard “left home in his pajamas.” Well, that’s a good summary of where Stability AI is.
The high school science club finds itself at odds with a mere school principal. The science club student knows that if the principal were capable, he would not be a mere principal. Thanks, MSFT Copilot. Were your senior managers in a high school science club?
The write up points out that Stability was the progenitor of Stable Diffusion, the art generator. I noticed the psycho-babbly terms stability and stable. Did you? Did the investors? Did the employees? Answer: Hey, there’s money to be made.
I noted this statement in the article:
The collaborative relationship between the investors and the promising startup gradually morphed into something more akin to that of a parent and an unruly child as the extent of internal turmoil and lack of clear direction at Stability became apparent, and even increased as Stability used its funding to expand its ranks.
Yep, high school management methods: “Don’t tell me what to do. I am smarter than you, Mr. Assistant Principal. You need me on the Quick Recall team, so go away,” echo in my mind in an Ezoic AI voice.
The write up continued the tale of mismanagement and adolescent angst, quoting the founder of Stability AI:
“Nobody tells you how hard it is to be a CEO and there are better CEOs than me to scale a business,” Mostaque said. “I am not sure anyone else would have been able to build and grow the research team to build the best and most widely used models out there and I’m very proud of the team there. I look forward to moving onto the next problem to handle and hopefully move the needle.”
I interpreted this as, “I did not know that calcium carbide in the lab sink drain could explode when in contact with water and then ignited, Mr. Principal.”
And, finally, let me point out this statement:
Though Stability AI’s models can still generate images of space unicorns and Lego burgers, music, and videos, the company’s chances of long-term success are nothing like they once appeared. “It’s definitely not gonna make me rich,” the investor says.
Several observations:
- Stability may presage the future for other high-flying and low-performing AI outfits. Why? Because teen management skills are problematic in a so-so economic environment
- AI is everywhere and its value is now derived by having something that solves a problem people will pay to have ameliorated. Shiny stuff fresh from the lab won’t make stakeholders happy
- Discipline, particularly in high school science club members, may not be what a dinobaby like me would call rigorous. Sloppiness produces a mess and lost opportunities.
Net net: Ask about a potential employer’s high school science club memories.
Stephen E Arnold, April 4, 2024
Publishers and Libraries: Tensions Escalate
April 4, 2024
This essay is the work of a dumb dinobaby. No smart software required.
We doubt this is what Ben Franklin had in mind. With more and more readers turning to digital editions, ABC News reports, “Libraries Struggle to Afford the Demand for E-Books, Seek New State Laws in Fight with Publishers.” With physical books, the process of building a library collection is simple: a volume is purchased (or donated) then loaned out repeatedly until it is lost or disintegrates. But publishers have made the process for ebooks much more complicated. And costly. Journalist Susan Haigh writes:
“The digital titles often come with a price tag that’s far higher than what consumers pay. While one hardcover copy of [Robin] Cook’s latest novel costs the library $18, it costs $55 to lease a digital copy — a price that can’t be haggled with publishers. And for that, the e-book expires after a limited time, usually after one or two years, or after 26 checkouts, whichever comes first. While e-books purchased by consumers can last into perpetuity, libraries need to renew their leased e-material. The modestly funded West Haven Library has spent more than $12,000 over the last three years to lease just 276 additional digital titles beyond what patrons can access through a consortium of public libraries. Eighty-four of those books are no longer available. If that same amount had been spent on paper books, it would have covered about 800 titles. … Publishers, however, argue the arrangement is fair considering e-book licenses for libraries allow numerous patrons to ‘borrow’ them and the per-reader cost is much less expensive than the per-reader rate.”
Well,yes, that is how public libraries work. Or it used to be. Will publishers come for hard copies next? Librarians across the US are pushing for legislation to counter these trends, and bills have been proposed in several states. Any that get passed, though, will have to make it through Big Publishing’s legal challenges. See the write-up for some lawmakers’ strategies to do so. Will libraries, and the taxpayers that fund them, prevail over these corporations? Stay tuned.
Cynthia Murrell, April 4, 2024
Preligens: An Important French AI Intelware Vendor May Be for Sale
April 3, 2024
This essay is the work of a dumb dinobaby. No smart software required.
I profiled Preligens (formerly Earthcube), the French specialized software firm with quite remarkable smart software, in one of my lectures a couple of years ago. Preligens processes satellite imagery and uses its home-brew AI system to identify objects. When I was in Paris last year, I spoke with some of my former colleagues at Exalead (now a unit of Dassault Systèmes), acquaintances from my pre-retirement travels, and some individuals I met online. I picked up a couple of rumors. One was that Preligens had tuned its system to monitor the license tags and vehicle models of cars, busses, and trucks. When a vehicle made too many passes in front of a structure of interest, Preligens’ AI would note that event and send an alert. I am reluctant to include the screenshots of the capabilities of the Preligens’ system. When I presented information about the company at my law enforcement lectures, several people investigating big-money yachts asked for the company’s Web site. I could not provide a point of contact because one of Preligens’ sales professionals replied to me via email and then disappeared. Oh, well.
Thanks, MSFT Copilot. I asked for lights from the corner window. But no, MSFT knows best. So good enough.
Why am I mentioning a French outfit founded in 2016 when the buzz is emanating from Mistral, a hot AI startup?
One of the items of unsubstantiated information I picked up was that the company needed money, and it was for sale. I spotted “Preligens Announces Surrender And Issues Call For Bids For Acquisition” in one of my feeds. The write seemed to corroborate what I heard as rumor in Paris; namely, the company is for sale. The write up says in what appears to be machine-translated French:
…the founders of Preligens, Arnaud Guérin and Renaud Allioux, turned to Jean-Yves Courtois last year – appointing him president of the company – in the hope of turning things around….The echoes reports that Jean-Yves Courtois has launched a call for tenders from around twenty players for its takeover and hopes for tender submissions in mid-April. Thales and Safran also seem to have entered the race.
The challenge for Preligens is that the company is tightly bound to the French military and it is going to consummate a deal unless the buyer is an outfit which passes the scrutiny of the French bureaucracy. As one US government agency learned a couple of years ago, Preligens would not sell all or part of the company to a US buyer. The Franco-American kumbaya sounds good, but when it comes to high-value AI technology, the progress of the discussions moved like traffic around the Arc de Triomphe right after Bastille Day. (You absolutely must watch the Légion étrangère troop. Magnificent, slow, and a reminder that one does not fool around with dudes wearing aprons and kepis.)
A deal can be crafted, but it will take work. The Preligens’ AI system is outstanding and extensible to a number of intelware and policeware use cases. There are some videos on YouTube plus the firm’s Web site if you want more information. The military-oriented information is not on those public sources. If you see me at an appropriate conference, I may let you look through my presentation about identifying submarine pens in an area quite close to a US friendly nation. Oh, the submarine pen was previously unknown prior to Preligens’ smart software knitting together data from satellite imagery. That is impressive, but the system was able to estimate the size of the pen. Very cool.
Stephen E Arnold, April 3, 2024
Angling to Land the Big Google Fish: A Humblebrag Quest to Be CEO?
April 3, 2024
This essay is the work of a dumb dinobaby. No smart software required.
My goodness, the staff and alums of DeepMind have been in the news. Wherever there are big bucks or big buzz opportunities, one will find the DeepMind marketing machinery. Consider “Can Demis Hassabis Save Google?” The headline has two messages for me. The first is that a “real” journalist things that Google is in big trouble. Big trouble translates to stakeholder discontent. That discontent means it is time to roll in a new Top Dog. I love poohbahing. But opining that the Google is in trouble. Sure, it was aced by the Microsoft-OpenAI play not too long ago. But the Softies have moved forward with the Mistral deal and the mysterious Inflection deal . But the Google has money, market share, and might. Jake Paul can say he wants the Mike Tyson death stare. But that’s an opinion until Mr. Tyson hits Mr. Paul in the face.
The second message in the headline that one of the DeepMind tribe can take over Google, defeat Microsoft, generate new revenues, avoid regulatory purgatory, and dodge the pain of its swinging door approach to online advertising revenue generation; that is, people pay to get in, people pay to get out, and soon will have to subscribe to watch those entering and exiting the company’s advertising machine.
Thanks, MSFT Copilot. Nice fish.
What are the points of the essay which caught my attention other than the headline for those clued in to the Silicon Valley approach to “real” news? Let me highlight a few points.
First, here’s a quote from the write up:
Late on chatbots, rife with naming confusing, and with an embarrassing image generation fiasco just in the rearview mirror, the path forward won’t be simple. But Hassabis has a chance to fix it. To those who known him, have worked alongside him, and still do — all of whom I’ve spoken with for this story — Hassabis just might be the perfect person for the job. “We’re very good at inventing new breakthroughs,” Hassabis tells me. “I think we’ll be the ones at the forefront of doing that again in the future.”
Is the past a predictor of future success? More than lab-to-Android is going to be required. But the evaluation of the “good at inventing new breakthroughs” is an assertion. Google has been in the me-too business for a long time. The company sees itself as a modern Bell Labs and PARC. I think that the company’s perception of itself, its culture, and the comments of its senior executives suggest that the derivative nature of Google is neither remembered nor considered. It’s just “we’re very good.” Sure “we” are.
Second, I noted this statement:
Ironically, a breakthrough within Google — called the transformer model — led to the real leap. OpenAI used transformers to build its GPT models, which eventually powered ChatGPT. Its generative ‘large language’ models employed a form of training called “self-supervised learning,” focused on predicting patterns, and not understanding their environments, as AlphaGo did. OpenAI’s generative models were clueless about the physical world they inhabited, making them a dubious path toward human level intelligence, but would still become extremely powerful. Within DeepMind, generative models weren’t taken seriously enough, according to those inside, perhaps because they didn’t align with Hassabis’s AGI priority, and weren’t close to reinforcement learning. Whatever the rationale, DeepMind fell behind in a key area.
Google figured something out and then did nothing with the “insight.” There were research papers and chatter. But OpenAI (powered in part by Sam AI-Man) used the Google invention and used it to carpet bomb, mine, and set on fire Google’s presumed lead in anything related to search, retrieval, and smart software. The aftermath of the Microsoft OpenAI PR coup is a continuing story of rehabilitation. From what I have seen, Google needs more time getting its ageingbody parts working again. The ad machine produces money, but the company reels from management issue to management issue with alarming frequency. Biased models complement spats with employees. Silicon Valley chutzpah causes neurological spasms among US and EU regulators. Something is broken, and I am not sure a person from inside the company has the perspective, knowledge, and management skills to fix an increasingly peculiar outfit. (Yes, I am thinking of ethnically-incorrect German soldiers loyal to a certain entity on Google’s list of questionable words and phrases.)
And, lastly, let’s look at this statement in the essay:
Many of those who know Hassabis pine for him to become the next CEO, saying so in their conversations with me. But they may have to hold their breath. “I haven’t heard that myself,” Hassabis says after I bring up the CEO talk. He instantly points to how busy he is with research, how much invention is just ahead, and how much he wants to be part of it. Perhaps, given the stakes, that’s right where Google needs him. “I can do management,” he says, ”but it’s not my passion. Put it that way. I always try to optimize for the research and the science.”
I wonder why the author of the essay does not query Jeff Dean, the former head of a big AI unit in Mother Google’s inner sanctum about Mr. Hassabis? How about querying Mr. Hassabis’ co-founder of DeepMind about Mr. Hassabis’ temperament and decision-making method? What about chasing down former employees of DeepMind and getting those wizards’ perspective on what DeepMind can and cannot accomplish.
Net net: Somewhere in the little-understood universe of big technology, there is an invisible hand pointing at DeepMind and making sure the company appears in scientific publications, the trade press, peer reviewed journals, and LinkedIn funded content. Determining what’s self-delusion, fact, and PR wordsmithing is quite difficult.
Google may need some help. To be frank, I am not sure anyone in the Google starting line up can do the job. I am also not certain that a blue chip consulting firm can do much either. Google, after a quarter century of zero effective regulation, has become larger than most government agencies. Its institutional mythos creates dozens of delusional Ulysses who cannot separate fantasies of the lotus eaters from the gritty reality of the company as one of the contributors to the problems facing youth, smaller businesses, governments, and cultural norms.
Google is Googley. It will resist change.
Stephen E Arnold, April 3, 2024
-
- Subscribe to Beyond Search
Feature archive
News archive
-
