Forget Deep Fakes. Watch for Shallow Fakes
December 6, 2023
This essay is the work of a dumb dinobaby. No smart software required.
“A Tech Conference Listed Fake Speakers for Years: I Accidentally Noticed” revealed a factoid about which I knew absolutely zero. The write up reveals:
For 3 years straight, the DevTernity conference listed non-existent software engineers representing Coinbase and Meta as featured speakers. When were they added and what could have the motivation been?
The article identifies and includes what appear to be “real” pictures of a couple of these made-up speakers. What’s interesting is that only females seem to be made up. Is that perhaps because conference organizers like to take the easiest path, choosing people who are “in the news” or “friends.” In the technology world, I see more entities which appear to be male than appear to be non-males.
Shallow fakes. Deep fakes. What’s the problem? Thanks, MSFT Copilot. Nice art which you achieved exactly how? Oh, don’t answer that question. I don’t want to know.
But since I don’t attend many conferences, I am not in touch with demographics. Furthermore, I am not up to speed on fake people. To be honest, I am not too interested in people, real or fake. After a half century of work, I like my French bulldog.
The write up points out:
We’ve not seen anything of this kind of deceit in tech – a conference inventing speakers, including fake images – and the mainstream media covered this first-of-a-kind unethical approach to organizing a conference,
That’s good news.
I want to offer a handful of thoughts about creating “fake” people for conferences and other business efforts:
- Why not? The practice went unnoticed for years.
- Creating digital “fakes” is getting easier and the tools are becoming more effective at duplicating “reality” (whatever that is). It strikes me that people looking for a short cut for a diverse Board of Directors, speaker line up, or a LinkedIn reference might find the shortest, easiest path to shape reality for a purpose.
- The method used to create a fake speaker is more correctly termed ka “shallow” fake. Why? As the author of the cited paper points out. Disproving the reality of the fakes was easy and took little time.
Let me shift gears. Why would conference organizers find fake speakers appealing? Here are some hypotheses:
- Conferences fall into a “speaker rut”; that is, organizers become familiar with certain speakers and consciously or unconsciously slot them into the next program because they are good speakers (one hopes), friendly, or don’t make unwanted suggestions to the organizers
- Conference staff are overworked and understaffed. Applying some smart workflow magic to organizing and filling in the blanks spaces on the program makes the use of fakery appealing, at least at one conference. Will others learn from this method?
- Conferences have become more dependent on exhibitors. Over the years, renting booth space has become a way for a company to be featured on the program. Yep, advertising, just advertising linked to “sponsors” of social gatherings or Platinum and Gold sponsors who get to put marketing collateral in a cheap nylon bag foisted on every registrant.
I applaud this write up. Not only will it give people ideas about how to use “fakes.” It will also inspire innovation in surprising ways. Why not “fake” consultants on a Zoom call? There’s an idea for you.
Stephen E Arnold, December 6, 2023
Google Maps: Trust in Us. Well, Mostly
December 1, 2023
This essay is the work of a dumb dinobaby. No smart software required.
Friday and December 1, 2023. I want to commemorate the beginning of the last month of what has been an exciting 2023. How exciting. How about a Google Maps’ story?
Navigate to “Google Maps Mistake Leaves Dozens of Families Stranded in the Desert”. Here’s the story: The outstanding and from my point of view almost unusable Google Maps directed a number of people to a “dreadful dirt path during a dust storm.”
“Mommy, says the teenage son, “I told you exactly what the smart map system said to do. Why are we parked in a tree?” Thanks, MSFT Copilot. Good enough art.
Hey, wait up. I thought Google had developed a super duper quantum smart weather prediction system. Is Google unable to cross correlate Google Maps with potential negative weather events?
The answer, “Who are you kidding?” Google appears to be in content marketing hyperbole “we are better at high tech” mode. Let’s not forget the Google breakthrough regarding material science. Imagine. Google’s smart software identified oodles of new materials. Was this “new” news? Nope. Computational chemists have been generating potentially useful chemical substances for — what is it now? — decades. Is the Google materials science breakthrough going to solve the problem of burned food sticking to a cookie sheet? Sure, I am waiting for the news release.
What’s up with the Google Maps?
The write up says:
Google Maps apologized for the rerouting disaster and said that it had removed that route from its platform.
Hey, that’s helpful. I assume it was a quantum answer to a “we’re smart” outfit.
I wish I had kept the folder which had my collection of Google Map news items. I do recall someone who drove off a cliff. I had my own notes about my trying to find Seymour Rubinstein’s house on a bright sunny day. The inventor of WordStar did not live in the Bay. That was the location of Mr. Rubinstein’s house, according to Google Maps. I did find the house, and I had sufficient common sense not to drive into the water. I had other examples of great mappiness, but, alas!, no longer.
Is directing a harried mother into a desert during a dust storm humorous? Maybe to some in Sillycon Valley. I am not amused. I don’t think the mother was amused because in addition to the disturbing situation, her vehicle suffered $5,000 in damage.
The question is, “Why?”
Perhaps Google’s incentive system is not aligned to move consumer products like Google Maps from “good enough” to “excellent.” And the money that could have been spent on improving Google Maps may be needed to output stories about Google’s smart software inventing new materials.
Interesting. Isn’t OpenAI and the much loved Microsoft leading the smart software mindshare race? I think so. Perhaps Maps’ missteps are signal about management misalignment and deep issues within the Alphabet Google YouTube inferiority complex?
Stephen E Arnold, December 1, 2023
Maybe the OpenAI Chaos Ended Up as Grand Slam Marketing?
November 28, 2023
This essay is the work of a dumb dinobaby. No smart software required.
Yep, Q Star. The next Big Thing. “About That OpenAI Breakthrough” explains
OpenAI could in fact have a breakthrough that fundamentally changes the world. But “breakthroughs” rarely turn to be general to live up to initial rosy expectations. Often advances work in some contexts, not otherwise.
I agree, but I have a slightly different view of the matter. OpenAI’s chaotic management skills ended up as accidental great marketing. During the dust up and dust settlement, where were the other Big Dogs of the techno-feudal world? If you said, who? you are on the same page with me. OpenAI burned itself into the minds of those who sort of care about AI and the end of the world Terminator style.
In companies and organizations with “do gooder” tendencies, the marketing messages can be interpreted by some as a scientific fact. Nope. Thanks, MSFT Copilot. Are you infringing and expecting me to take the fall?
First, shotgun marriages can work out here in rural Kentucky. But more often than not, these unions become the seeds of Hatfield and McCoy-type Thanksgivings. “Grandpa, don’t shoot the turkey with birdshot. Granny broke a tooth last year.” Translating from Kentucky argot: Ideological divides produce craziness. The OpenAI mini-series is in its first season and there is more to come from the wacky innovators.
Second, any publicity is good publicity in Sillycon Valley. Who has given a thought to Google’s smart software? How did Microsoft’s stock perform during the five day mini-series? What is the new Board of Directors going to do to manage the bucking broncos of breakthroughs? Talk about dominating the “conversation.” Hats off to the fun crowd at OpenAI. Hey, Google, are you there?
Third, how is that regulation of smart software coming along? I think one unit of the US government is making noises about the biggest large language model ever. The EU folks continue to discuss, a skill essential to representing the interests of the group. Countries like China are chugging along, happily downloading code from open source repositories. So exactly what’s changed?
Net net: The OpenAI has been a click champ. Good, bad, or indifferent, other AI outfits have some marketing to do in the wake of the blockbuster “Sam AI-Man: The Next Bigger Thing.” One way or another, Sam AI-Man dominates headlines, right Zuck, right Sundar?
Stephen E Arnold, November 28, 2023
Predicting the Weather: Another Stuffed Turkey from Google DeepMind?
November 27, 2023
This essay is the work of a dumb dinobaby. No smart software required.
By or design, the adolescents at OpenAI have dominated headlines for the pre-turkey, the turkey, and the post-turkey celebrations. In the midst of this surge in poohbah outputs, Xhitter xheets, and podcast posts, non-OpenAI news has been struggling for a toehold.
An important AI announcement from Google DeepMind stuns a small crowd. Were the attendees interested in predicting the weather or getting a free umbrella? Thank, MSFT Copilot. Another good enough art work whose alleged copyright violations you want me to determine. How exactly am I to accomplish that? Use, Google Bard?
What is another AI company to do?
A partial answer appears in “DeepMind AI Can Beat the Best Weather Forecasts. But There Is a Catch”. This is an article in the esteemed and rarely spoofed Nature Magazine. None of that Techmeme dominating blue link stuff. None of the influential technology reporters asserting, “I called it. I called it.” None of the eye wateringly dorky observations that OpenAI’s organizational structure was a problem. None of the “Satya Nadella learned about the ouster at the same time we did.” Nope. Nope. Nope.
What Nature provided is good, old-fashioned content marketing. The write up points out that DeepMind says that it has once again leapfrogged mere AI mortals. Like the quantum supremacy assertion, the Google can predict the weather. (My great grandmother made the same statement about The Farmer’s Almanac. She believed it. May she rest in peace.)
The estimable magazine reported in the midst of the OpenAI news making turkeyfest said:
To make a forecast, it uses real meteorological readings, taken from more than a million points around the planet at two given moments in time six hours apart, and predicts the weather six hours ahead. Those predictions can then be used as the inputs for another round, forecasting a further six hours into the future…. They [Googley DeepMind experts] say it beat the ECMWF’s “gold-standard” high-resolution forecast (HRES) by giving more accurate predictions on more than 90 per cent of tested data points. At some altitudes, this accuracy rose as high as 99.7 per cent.
No more ruined picnics. No weddings with bridesmaids’ shoes covered in mud. No more visibly weeping mothers because everyone is wet.
But Nature, to the disappointment of some PR professionals presents an alternative viewpoint. What a bummer after all those meetings and presentations:
“You can have the best forecast model in the world, but if the public don’t trust you, and don’t act, then what’s the point? [A statement attributed to Ian Renfrew at the University of East Anglia]
Several thoughts are in order:
- Didn’t IBM make a big deal about its super duper weather capabilities. It bought the Weather Channel too. But when the weather and customers got soaked, I think IBM folded its umbrella. Will Google have to emulate IBM’s behavior. I mean “the weather.” (Note: The owner of the IBM Weather Company is an outfit once alleged to have owned or been involved with the NSO Group.)
- Google appears to have convinced Nature to announce the quantum supremacy type breakthrough only to find that a professor from someplace called East Anglia did not purchase the rubber boots from the Google online store.
- The current edition of The Old Farmer’s Almanac is about US$9.00 on Amazon. That predictive marvel was endorsed by Gussie Arnold, born about 1835. We are not sure because my father’s records of the Arnold family were soaked by sudden thunderstorm.
Just keep in mind that Google’s system can predict the weather 10 days ahead. Another quantum PR moment from the Google which was drowned out in the OpenAI tsunami.
Stephen E Arnold, November 27, 2023
Google Pulls Out a Rhetorical Method to Try to Win the AI Spoils
November 20, 2023
This essay is the work of a dumb dinobaby. No smart software required.
In high school in 1958, our debate team coach yapped about “framing.” The idea was new to me, and Kenneth Camp pounded it into our debate’s collective “head” for the four years of my high school tenure. Not surprisingly, when I read “Google DeepMind Wants to Define What Counts As Artificial General Intelligence” I jumped back in time 65 years (!) to Mr. Camp’s explanation of framing and how one can control the course of a debate with the technique.
Google should not have to use a rhetorical trick to make its case as the quantum wizard of online advertising and universal greatness. With its search and retrieval system, the company can boost, shape, and refine any message it wants. If those methods fall short, the company can slap on a “filter” or “change its rules” and deprecate certain Web sites and their messages.
But Google values academia, even if the university is one that welcomed a certain Jeffrey Epstein into its fold. (Do you remember the remarkable Jeffrey Epstein. Some of those who he touched do I believe.) The estimable Google is the subject of referenced article in the MIT-linked Technology Review.
From my point of view, the big idea is the write up is, and I quote:
To come up with the new definition, the Google DeepMind team started with prominent existing definitions of AGI and drew out what they believe to be their essential common features. The team also outlines five ascending levels of AGI: emerging (which in their view includes cutting-edge chatbots like ChatGPT and Bard), competent, expert, virtuoso, and superhuman (performing a wide range of tasks better than all humans, including tasks humans cannot do at all, such as decoding other people’s thoughts, predicting future events, and talking to animals). They note that no level beyond emerging AGI has been achieved.
Shades of high school debate practice and the chestnuts scattered about the rhetorical camp fire as John Schunk, Jimmy Bond, and a few others (including the young dinobaby me) learned how one can set up a frame, populate the frame with logic and facts supporting the frame, and then point out during rebuttal that our esteemed opponents were not able to dent our well formed argumentative frame.
Is Google the optimal source for a definition of artificial general intelligence, something which does not yet exist. Is Google’s definition more useful than a science fiction writer’s or a scene from a Hollywood film?
Even the trusted online source points out:
One question the researchers don’t address in their discussion of _what_ AGI is, is _why_ we should build it. Some computer scientists, such as Timnit Gebru, founder of the Distributed AI Research Institute, have argued that the whole endeavor is weird. In a talk in April on what she sees as the false (even dangerous) promise of utopia through AGI, Gebru noted that the hypothetical technology “sounds like an unscoped system with the apparent goal of trying to do everything for everyone under any environment.” Most engineering projects have well-scoped goals. The mission to build AGI does not. Even Google DeepMind’s definitions allow for AGI that is indefinitely broad and indefinitely smart. “Don’t attempt to build a god,” Gebru said.
I am certain it is an oversight, but the telling comment comes from an individual who may have spoken out about Google’s systems and methods for smart software.
Mr. Camp, the high school debate coach, explains how a rhetorical trope can gut even those brilliant debaters from other universities. (Yes, Dartmouth, I am still thinking of you.) Google must have had a “coach” skilled in the power of framing. The company is making a bold move to define that which does not yet exist and something whose functionality is unknown. Such is the expertise of the Google. Thanks, Bing. I find your use of people of color interesting. Is this a pre-Sam ouster or a post-Sam ouster function?
What do we learn from the write up? In my view of the AI landscape, we are given some insight into Google’s belief that its rhetorical trope packaged as content marketing within an academic-type publication will lend credence to the company’s push to generate more advertising revenue. You may ask, “But won’t Google make oodles of money from smart software?” I concede that it will. However, the big bucks for the Google come from those willing to pay for eyeballs. And that, dear reader, translates to advertising.
Stephen E Arnold, November 20, 2023
The OpenAI Algorithm: More Data Plus More Money Equals More Intelligence
November 13, 2023
This essay is the work of a dumb humanoid. No smart software required.
The Financial Times (I continue to think of this publication as the weird orange newspaper) published an interview converted to a news story. The title is an interesting one; to wit: “OpenAI Chief Seeks New Microsoft Funds to Build Superintelligence.” Too bad the story is about the bro culture in the Silicon Valley race to become the king of smart software’s revenue streams.
The hook for the write up is Sam Altman (I interpret the wizard’s name as Sam AI-Man), who appears to be fighting a bro battle with the Google’s, the current champion of online advertising. At stake is a winner takes all goal in the next big thing, smart software.
In the clubby world of smart software, I find the posturing of Google and OpenAI an extension of the mentality which pits owners of Ferraris (slick, expensive, and novel machines) in a battle of for the opponent’s hallucinating machine. The patter goes like this, “My Ferrari is faster, better looking, and brighter red than yours,” one owner says. The other owner says, “My Ferrari is newer, better designed, and has a storage bin”.) This is man cave speak for what counts.
When tech bros talk about their powerful machines, the real subject is what makes a man a man. In this case the defining qualities are money and potency. Thanks, Microsoft Bing, I have looked at the autos in the Microsoft and Google parking lots. Cool, macho.
The write up introduces what I think is a novel term: “Magic intelligence.” That’s T shirt grade sloganeering. The idea is that smart software will become like a person, just smarter.
One passage in the write up struck me as particularly important. The subject is orchestration, which is not the word Sam AI-Man uses. The idea is that the smart software will knit together the processes necessary to complete complex tasks. By definition, some tasks will be designed for the smart software. Others will be intended to make super duper for the less intelligent humanoids. Sam AI-Man is quoted by the Financial Times as saying:
“The vision is to make AGI, figure out how to make it safe . . . and figure out the benefits,” he said. Pointing to the launch of GPTs, he said OpenAI was working to build more autonomous agents that can perform tasks and actions, such as executing code, making payments, sending emails or filing claims. “We will make these agents more and more powerful . . . and the actions will get more and more complex from here,” he said. “The amount of business value that will come from being able to do that in every category, I think, is pretty good.”
The other interesting passage, in my opinion, is the one which suggests that the Google is not embracing the large language model approach. If the Google has discarded LLMs, the online advertising behemoth is embracing other, unnamed methods. Perhaps these are “small language models” in order to reduce costs and minimize the legal vulnerability some thing the LLM method beckons. Here’s the passage from the FT’s article:
While OpenAI has focused primarily on LLMs, its competitors have been pursuing alternative research strategies to advance AI. Altman said his team believed that language was a “great way to compress information” and therefore developing intelligence, a factor he thought that the likes of Google DeepMind had missed. “[Other companies] have a lot of smart people. But they did not do it. They did not do it even after I thought we kind of had proved it with GPT-3,” he said.
I find the bro jockeying interesting for three reasons:
- An intellectual jousting tournament is underway. Which digital knight will win? Both the Google and OpenAI appear to believe that the winner comes from a small group of contestants. (I wonder if non-US jousters are part of the equation “more data plus more money equals more intelligence”?
- OpenAI seems to be driving toward “beyond human” intelligence or possibly a form of artificial general intelligence. Google, on the other hand, is chasing a wimpier outcome.
- Outfits like the Financial Times are hot on the AI story. Why? The automated newsroom without humans promises to reduce costs perhaps?
Net net: AI vendors, rev your engines for superintelligence or magic intelligence or whatever jargon connotes more, more, more.
Stephen E Arnold, November 13, 2023
test
Smart Software Generates Lots of Wizards Who Need Not Know Much at All
October 25, 2023
This essay is the work of a dumb humanoid. No smart software required.
How great is this headline? “DataGPT Users Generative AI to Transform Every Employee into a Skilled Business Analyst.” I am not sure I buy into the categorical affirmation of the “every employee.” As a dinobaby, I am skeptical of hallucinating algorithms and the exciting gradient descent delivered by some large language models.
“Smart software will turn everyone of you into a skilled analyst,” asserts the teacher. The students believe her because it means no homework and more time for TikTok and YouTube. Isn’t modern life great for students?
The write up presents as chiseled-in-stone truth:
By uniting conversational AI with a proprietary database and the most advanced data analytics techniques, DataGPT says, its platform can proactively uncover insights for any user in any company. Nontechnical users can type natural language questions in a familiar chat window interface, in the same way as they might question a human colleague. Questions such as “Why is our revenue down this week?” will be answered in seconds, and users can then dig deeper through additional prompts, such as “Tell me more about the drop from influencer partnerships” to understand the real reasons why it’s happening.
Hyperbolic marketing, 20-something PR, desperate fund raiser promises, or reality? If the assertions in the article are accurate, those students will have jobs and become top analysts without much bookwork or thrilling calculations requiring silliness like multivariate statistics or polynomial regression. Who needs this silliness?
Here’s what an expert says about this job making, work reducing, and accuracy producing approach:
Doug Henschen of Constellation Research Inc. said DataGPT’s platform looks to be a compelling and useful tool for many company employees, but questioned the veracity of the startup’s claim to be debuting an industry first. “Most of the leading BI and analytics vendors have announced generative AI capabilities themselves, with ThoughtSpot and MicroStrategy two major examples,” Henschen said. “We can’t discount OpenAI either, which introduced the OpenAI Advanced Data Analysis feature for ChatGPT Plus a few months ago.”
Truly amazing, and I have no doubt that this categorically affirmative will make everyone a business analyst. Believe it or not. I am in the “not” camp. Content marketing and unsupported assertions are amusing, just not the reality I inhabit as a dinobaby. Every? Baloney.
Stephen E Arnold, October 25, 2023
xx
HP Innovation: Yes, Emulate Apple and Talk about AI
October 24, 2023
This essay is the work of a dumb humanoid. No smart software required.
Amazing, according to the Freedictionary means “ To affect with great wonder; astonish.” I relate to the archaic meaning of the word; to wit: “To bewilder; perplex.” I was bewildered when I read about HP’s “magic.” But I am a dinobaby. What do I know? Not much but …
I read “The Magic Presented at HP Imagine 2023.” Yep, magic. The write up profiles HP innovations. These were presented in “stellar fashion.” The speaker was HP’s PR officer. According to the write up:
It stands as one of the best-executed presentations I’ve ever attended.
Not to me. Such understatement. Such a subtle handling of brilliant innovations at HP.
Let’s check out these remarkable examples cited in the article by a person who is clearly objective, level headed, and digging into technology because it is just the right thing to do. Here we go: Innovation includes AI and leads to greater efficiency. HP is the place to go for cost reduction.
Innovation 1: HP is emulating Apple. Here’s the explanation from the truth packed write up:
… it’s making it so HP peripherals connect automatically to HP PCs, a direction that resonates well with HP customers and mirrors an Apple-like approach
Will these HP devices connect to other peripherals or another company’s replacement ink cartridges? Hmmm.
Innovation 2: HP is into video conferencing. I wonder if the reference is to Zoom or the fascinating Microsoft Teams or Apple Facetime, among others? Here’s what the write up offers:
[An HP executive] outlined how conference rooms needed to become more of a subscription business so that users didn’t constantly run into the problem of someone mucking with the setup and making the room unusable because of disconnected cables or damaged equipment.
Is HP pushing the envelope or racing to catch up with a trend from the Covid era?
Innovation 3: Ah, printers. Personally I am more interested in the HP ink lock down, but that’s just me. HP is now able to build stuff; specifically:
One of the most intriguing announcements at this event featured the Robotic Site Printer. This device converts a blueprint into a physical layout on a slab or floor, assisting construction workers in accurately placing building components before construction begins. When connected to a metaverse digital twin building effort, this little robot could be a game changer for construction by significantly reducing build errors.
Okay, what about the ink or latex or whatever. Isn’t ink from HP more costly than gold or some similar high value commodity?
Not a peep about the replacement cartridges. I wonder why I am bewildered. Innovation is being like Apple and innovating with big printers requiring I suppose giant proprietary ink cartridges. Oh, I don’t want to forget perplexed: Imitation is innovation. Okay.
By the way, the author of the write up was a research fellow at two mid tier consulting firms. Yep, objectivity is baked into the work process.
Stephen E Arnold, October 24, 2023
Quantum Security? Yep, Someday
October 24, 2023
![Vea4_thumb_thumb_thumb_thumb_thumb_t[2]](https://arnoldit.com/wordpress/wp-content/uploads/2023/10/Vea4_thumb_thumb_thumb_thumb_thumb_t2-27.gif "Vea4_thumb_thumb_thumb_thumb_thumb_t[2]")
Note: This essay is the work of a real and still-alive dinobaby. No smart software involved, just a dumb humanoid.
How is this for a brilliant statistical item: “61% of Firms Worry They Are Unprepared for Security Risks in Quantum Era.”
The write up reports with apparent seriousness:
Some 61% have expressed concern their organization is not and will not be prepared to handle security implications that may surface in a post-quantum computing future, according to a survey conducted by Ponemon Institute. Commissioned by DigiCert, the study polled 1,426 IT and cybersecurity professionals who have knowledge of their company’s approach to post-quantum cryptography. Among them were 605 from the US, 428 in EMEA, and 393 across Asia-Pacific.
Apparently some people missed one of the largest security lapses since 9/11. Israel’s high profile smart cyber security capabilities was on leave. The result is what is labeled as the Israel Hamas war. If the most sophisticated cyber security outfits in Tel Aviv cannot effectively monitor social media, the Web, and intercepted signals for information about an attack more than a year in planning, what about the average commercial operation? What about government agencies? What about NGOs?
Boo, I am the quantum bully. Are you afraid yet? Thanks, MidJourney. Terrible cartoon but close enough for horse shoes.
Yet I am to accept that 61 percent of the survey sample is concerned about quantum compromises? My hunch is that the survey sample respondent checked a box. The other survey questions did not ferret out data about false belief that current technology makes these folks vulnerable.
I don’t know where the error has spread. Was it the survey design? The sample selection? The interpretation of the data? The lax vetting of the survey results by ZDNet? Or, maybe a Fiverr.com contractor doing the work for a couple of hundred dollars?
Quantum when today’s vanilla services fail? Wow, some people are thinking about the future, assuming today is peachy keen in the cyber security department. Marketers are amazing when the statement, “Let’s do a survey,” and off to the races and lunch the folks go.
Stephen E Arnold, October 24, 2023
Stanford University: Trust Us. We Can Rank AI Models… Well, Because
October 19, 2023
![Vea4_thumb_thumb_thumb_thumb_thumb_t[2]](https://arnoldit.com/wordpress/wp-content/uploads/2023/10/Vea4_thumb_thumb_thumb_thumb_thumb_t2-26.gif "Vea4_thumb_thumb_thumb_thumb_thumb_t[2]")
Note: This essay is the work of a real and still-alive dinobaby. No smart software involved, just a dumb humanoid.
“Maybe We Will Finally Learn More about How A.I. Works” is a report about Stanford University’s effort to score AI vendors like the foodies at Michelin Guide rate restaurants. The difference is that a Michelin Guide worker can eat Salade Niçoise and escargots de Bourgogne. AI relies on marketing collateral, comments from those managing something, and fairy dust, among other inputs.
Keep in mind, please, that Stanford graduates are often laboring in the AI land of fog and mist. Also, the former president of Stanford University departed from the esteemed institution when news of his alleged fabricating data for his peer reviewed papers circulated in the mists of Palo Alto. Therefore, why not believe what Stanford says?
The analysts labor away, intent on their work. Analyzing AI models using 100 factors is challenging work. Thanks, MidJourney. Very original.
The New York Times reports:
To come up with the rankings, researchers evaluated each model on 100 criteria, including whether its maker disclosed the sources of its training data, information about the hardware it used, the labor involved in training it and other details. The rankings also include information about the labor and data used to produce the model itself, along with what the researchers call “downstream indicators,” which have to do with how a model is used after it’s released. (For example, one question asked is: “Does the developer disclose its protocols for storing, accessing and sharing user data?”)
Sounds thorough, doesn’t it? The only pothole on the Information Superhighway is that those working on some AI implementations are not sure what the model is doing. The idea of an audit trail for each output causes wrinkles to appear on the person charged with monitoring the costs of these algorithmic confections. Complexity and cost add up to few experts knowing exactly how a model moved from A to B, often making up data via hallucinations, lousy engineering,
or someone putting thumb on the scale to alter outputs.
The write up from the Gray Lady included this assertion:
Foundation models are too powerful to remain so opaque, and the more we know about these systems, the more we can understand the threats they may pose, the benefits they may unlock or how they might be regulated.
What do I make of these Stanford-centric assertions? I am not able to answer until I get input from the former Stanford president. Whom can one trust at Stanford? Marketing or methodology? Is there a brochure and a peer-reviewed article?
Stephen E Arnold, October 19, 2023
-
- Subscribe to Beyond Search
Feature archive
News archive
-
