Goggle Points Out the ChatGPT Has a Core Neural Disorder: LSD or Spoiled Baloney?
February 16, 2023
I am an old-fashioned dinobaby. I have a reasonably good memory for great moments in search and retrieval. I recall when Danny Sullivan told me that search engine optimization improves relevance. In 2006, Prabhakar Raghavan on a conference call with a Managing Director of a so-so financial outfit explained that Yahoo had semantic technology that made Google’s pathetic effort look like outdated technology.
Hallucinating pizza courtesy of the super smart AI app Craiyon.com. The art, not the write up it accompanies, was created by smart software. The article is the work of the dinobaby, Stephen E Arnold. Looks like pizza to me. Close enough for horseshoes like so many zippy technologies.
Now that SEO and its spawn are scrambling to find a way to fiddle with increasingly weird methods for making software return results the search engine optimization crowd’s customers demand, Google’s head of search Prabhakar Raghavan is opining about the oh, so miserable work of Open AI and its now TikTok trend ChatGPT. May I remind you, gentle reader, that OpenAI availed itself of some Googley open source smart software and consulted with some Googlers as it ramped up to the tsunami of PR ripples? May I remind you that Microsoft said, “Yo, we’re putting some OpenAI goodies in PowerPoint.” The world rejoiced and Reddit plus Twitter kicked into rave mode.
Google responded with a nifty roll out in Paris. February is not April, but maybe it should have been in April 2023, not in les temp d’hiver?
I read with considerable amusement “Google Vice President Warns That AI Chatbots Are Hallucinating.” The write up states as rock solid George Washington I cannot tell a lie truth the following:
Speaking to German newspaper Welt am Sonntag, Raghavan warned that users may be delivered complete nonsense by chatbots, despite answers seeming coherent. “This type of artificial intelligence we’re talking about can sometimes lead to something we call hallucination,” Raghavan told Welt Am Sonntag. “This is then expressed in such a way that a machine delivers a convincing but completely fictitious answer.”
LSD or just the Google code relied upon? Was it the Googlers of whom OpenAI asked questions? Was it reading the gems of wisdom in Google patent documents? Was it coincidence?
I recall that Dr. Timnit Gebru and her co-authors of the Stochastic Parrot paper suggest that life on the Google island was not palm trees and friendly natives. Nope. Disagree with the Google and your future elsewhere awaits.
Now we have the hallucination issue. The implication is that smart software like Google-infused OpenAI is addled. It imagines things. It hallucinates. It is living in a fantasy land with bean bag chairs, Foosball tables, and memories of Odwalla juice.
I wrote about the after-the-fact yip yap from Google’s Chair Person of the Board. I mentioned the Father of the Darned Internet’s post ChatGPT PR blasts. Now we have the head of search’s observation about screwed up neural networks.
Yep, someone from Verity should know about flawed software. Yep, someone from Yahoo should be familiar with using PR to mask spectacular failure in search. Yep, someone from Google is definitely in a position to suggest that smart software may be somewhat unreliable because of fundamental flaws in the systems and methods implemented at Google and probably other outfits loving the Tensor T shirts.
Stephen E Arnold, February 16, 2023
Google Wizards: Hey, We Knew But Did Not Intervene. Very Bard Like
February 15, 2023
I read two stories. Each offers a glimpse into what I call backing away and distancing. I think each reveals the failure of Google governance. You may disagree. That’s okay, particularly if the stories are horse feathers. My hunch is that there is a genetically warped turkey under the plumage.
The first item is from the increasingly sensational Insider. The story is “Google Didn’t Think Its Bard AI Was Really Ready for a Product Yet, Says Alphabet Chairman, Days after Its Stock Fell Following the Chatbot’s Very Public Mistake.” The write up pivots on information (allegedly 100 percent dead solid in the bull’s eye) provided by John Hennessy, the chairman of Alphabet. The chair person! What did this captain of the digital titan say? I quote from the write up:
“I think Google was hesitant to productize this because it didn’t think it was really ready for a product yet, but, I think, as a demonstration vehicle, it’s a great piece of technology….He added Google was slow to introduce Bard because it was still giving wrong answers.
From my point of view, isn’t the role of the Board of Directors, and specifically the Chair, supposed to provide what might be called governance guidance? Since this admission of “giving wrong answers” is made public after the disaster in a city where a great lunch is easy to obtain, I would suggest that the bowl of soupe a l’oignon was prepared from a bag of instant convenient food: Not particularly good but perfect for a high school science club snack.
The second item is from CNet, which has some experience with smart software. The article is “Computing Guru Criticizes ChatGPT AI Tech for Making Things Up.” And who is the computing guru? None other than Vint Cerf, one of the father’s of the Internet if I remember something I heard at a conference.
The CNet article reported as actual factual:
But, speaking Monday [February 13, 2023] at Celesta Capital’s TechSurge Summit, he did warn about ethical issues of a technology that can generate plausible sounding but incorrect information even when trained on a foundation of factual material. If an executive tried to get him to apply ChatGPT to some business problem, his response would be to call it snake oil, referring to bogus medicines that quacks sold in the 1800s, he said. Another ChatGPT metaphor involved kitchen appliances.
Then this allegedly accurate quotation from the father of the Internet and Google guru:
“It’s like a salad shooter — you know how the lettuce goes all over everywhere,” Cerf said. “The facts are all over everywhere, and it mixes them together because it doesn’t know any better.”
Did the Googlers crafting Bard run the demonstration by Mr. Cerf? Nope. The write up says:
Cerf said he was surprised to learn that ChatGPT could fabricate bogus information from a factual foundation. “I asked it, ‘Write me a biography of Vint Cerf.’ It got a bunch of things wrong,” Cerf said. That’s when he learned the technology’s inner workings — that it uses statistical patterns spotted from huge amounts of training data to construct its response. “It knows how to string a sentence together that’s grammatically likely to be correct,” but it has no true knowledge of what it’s saying, Cerf said. “We are a long way away from the self-awareness we want.”
It seems to me that if the father of the Internet is on staff, it would make sense to get some inputs.
Let’s recap:
- After the fact, the Chair of the Board points out known problems but does not invoke action based on the need for governance related to product performance. Seems like something slipped betwixt the cup and the lip.l
- After the fact, the father of the Internet points out that he was “surprised” that Google technology generated misinformation. Again … after the fact.
Is the company managed by responsible adults or individuals who believe themselves to be in a high school science club? Are Googlers indifferent to the need to get their act together before they take the show on the road.
I think the French could label either Googlers’ comment as observations offered in l’esprit de l’escalier. Accurate but not management.
Stephen E Arnold, February 15, 2023
The Chinese Balloon: A Legacy of Loon?
February 14, 2023
Several of my monographs about the Google relied on text and link analysis of Google patent applications and patents. One of the patents was “Balloon Altitude Control Using Density Adjustment and or Volume Adjustment.” You may recall the Loon Balloon, a project to provide Internet access in locations where landlines or other delivery mechanisms were either not affordable, safe, or accessible due to an issue. (An “issue” could be a war, disease, or a populace with a leader who was un-Googley.)
What is the likelihood that China’s use of a Loon-like invention described in US9033274B2 coupled with some smart software has enabled the Middle Kingdom to enable global activities. Balloons can carry explosive devices, surveillance equipment, or electronics designed to screw up a range of electrical (wave) centric systems.
Probably just a random connection my brain generated. Probably nothing significant.
Stephen E Arnold, February 13, 2023
Prabhakar in Paris: An Expensive Google Trip
February 13, 2023
Paris has good restaurants, and it has quite a few alert, well-educated people. So why did Google take the Prabhakar Smart Search Show to the City of Light? “Google Employees Criticize CEO Sundar Pichai for Rushed, Botched Announcement of GPT Competitor Bard” does not have an answer for me or for others either.
The write up states:
Staffers took to the popular internal forum Memegen [an in house Google thing] to express their thoughts on the Bard announcement, referring to it as “rushed,” “botched” and “un-Googley,” according to messages and memes viewed by CNBC.
But here’s the killer comment:
During Google’s Wednesday event, search boss Prabhakar Raghavan briefly shared some slides with examples of Bard’s capabilities. People tuning in expected to hear more, and some employees weren’t even aware of the event. One presenter forgot to bring a phone that was required for the demo. Meanwhile, people on Twitter began pointing out that an ad for Bard offered an incorrect description of a telescope used to take the first pictures of a planet outside our solar system.
Is Prabhakar the Red Skelton of smart software infused search? By the way, the turning point for Googzilla was the interaction between the company and Dr. Timnit Gebru. If you have not read the stochastic parrot, you may find it interesting.
Polly want Google management to be organized? Squawk:
Dear Sundar, the Bard launch and the layoffs were rushed, botched, and myopic…. [now make parrot sounds]
The next high school reunion for Sundar and Prabhakar will be interesting indeed.
Stephen E Arnold, February 13, 2023
Google Shows Its Smarts by Trimming Its Market Value
February 10, 2023
The title of this blog is Beyond Search. More than a decade ago, I wanted to have a place to put my observations about search and retrieval. Retirement was coming, and I was unable to put criticism of search baloney in the write ups I was paid to do. (Nope, I won’t name the publication.)
Art generated and probably owned by Craiyon, Dreamtime, Getty, Alamy, Shutterstock and any other outfit looking to make a buck surfing on legal water droplets. I sure did not create this picture.
That’s why I have not been going head over heels with the smart software revolution. I now point to articles that offer something I find either interesting, amusing, or certifiably whacky. Today, I want to call your attention to a statement I quite like which appeared in “Google Bard or Google Storyteller”. Here’s the quote:
The problem here isn’t just the mistake. It’s the fact that this mistake was highlighted as an example of what Google Bard could accomplish. Before releasing this information, there were likely many people involved at Google. None were competent enough to fact-check what they wanted to show the world. This is not only embarrassing, but it also casts many doubts about Google’s internal checks on its products and shows an astounding level of amateurism for one of the biggest companies in the world.
Do you recall the antics of Abbott and Costello or the Three Stooges? I wonder if this slip betwixt cup and lip is the first program of the 2023 season for the Sundar and Prabhakar Show, sponsored by Microsoft and OpenAI where you just Bing it!
I can hear the announcer saying,
“With Jeff Dean, Marcus White, and special guest stars Larry Page and Sergey Brin. Here are Sundar and Prabhakar, who have just returned from a meeting at what’s left of Charlie’s Café where the talented duo were discussing smart search. We join Sundar and Prabhakar in the once glorious dining facility…”
What would the comedy script generated by Bard say? I don’t want to know because that loss in market value was a hoot appropriate for a thunder lizard with a broken leg in the snow.
Stephen E Arnold, February 10, 2023
Google Is Busy: What about YouTube Filtering?
February 10, 2023
We noted a reference to a video produced by Wendover Productions. I know zero about the outfit’s videos, but one caught my eye. The video is “How to Illegally Cro0ss the Mexico-US Border.” The video runs 14 minutes and has been online for more than a year. The company describes its “aboutness” this way:
Wendover Productions is all about explaining how our world works. From travel, to economics, to geography, to marketing and more, every video will leave you with a little better understanding of our world. New videos go out every other Tuesday.
The video does not strike me as particularly helpful. But there are some interesting factoids:
- Barriers protect about one-third of the border
- Walk across is possible in certain locations. Maybe Brownsville, Texas
- Cross in remote areas but a walk of 20 to 30 miles may be necessary
- Humanitarian groups have set up water stations in certain area
New methods of dealing with certain border infractions are in place and some like the Anduril tower and drone method seem to have some value. The situation, of course, determines what steps are taken and what methods are employed by authorized officials.
I wanted to highlight this video as I have those which provide information about to obtain and hack commercial software.
My question is, “Has Google been sufficiently distracted by bonuses, possible revenue shrinkage, and Code Red cartwheels to have ignored what appears to be information facilitating illegal activity.
Another possibility is that Google’s method of identifying certain types of content like stealing software and entering the US illegally is of little interest to Googlers or its smart software.
A related question is, “What will slip through the gaps in the Bard system?”
Stephen E Arnold, February 10, 2023
The Microsoft Vision: Agent-Intermediated Computing
February 9, 2023
Not one single input from smart software. — Editor
[Mise en scène] Googzilla is towering over advertisers, hissing and brandishing its long talons. Then the creature turns its head. It tiny ear slits twitch. He hears a sound like “boink” or “thunk.” Distracted, the 25 year-old terror looks from the cowering advertisers and fixes his maroon-hued eyes on a almost insignificant figure. That entity is Satya Nadella, who has just ruing Sundar’s and Prabhakar’s high school reunion. The fear of answering the question “Hey, how did you guys miss this ChatGPT – Microsoft thing” is terrifying. Googzilla emits a plaintive “welp.” The advertisers back away and start walking toward Redmond. [Fade to black]
The shoe has dropped. Boink or thunk, depending on your perceptual equipment. “Microsoft’s AI-Powered Bing Will Challenge Google Search” reports “Microsoft may finally have figured out how to get you to use Bing.” The article adds a quote directly from the champion of high school reunions in India:
“All computer interaction is going to be mediated with an agent helping,” Chief Executive Satya Nadella said at a launch event at the company’s headquarters in Redmond, Washington. “We’re going to have this notion of a co-pilot that’s going to be there across every application.”
I won’t point out that “all” is a logical impossibility when it comes to humanoids and computer systems. But I will let that slide… because ChatGPT.
Kudos to Microsoft for pulling off the marketing play of the year. I know it is only early February 2023, but it may take something truly special to tap into the brush fires ignited by ChatGPT. Will Bing be better? Maybe? Will the ChatGPT thing frighten the allegedly monopolistic Google? I think it has.
There are several examples that illustrate the disarray of the Google radar system. First, the search beacon missed the incoming ChatGPT balloon. Hello, Prabhakar. Isn’t and maybe wasn’t that your job?
Then there was the startling Code Red. Yeah, that’s professional. OpenAI has been around six or seven years. Now it is Code Red. Situational awareness seems to be lacking where Googzilla hunts. This is a flaw in an apex predator, is it not?
The dollop of whipped cream on this torched cupcake was asking the former head chefs of data hoovering and search engine optimization as an spur to buying advertising to return to the wizard lair. Yep, Sundar asked Sergey and Larry to help out with the Code Red thing. Okay, but let’s recall the origin of the Google money machine. Wasn’t it Yahoo-Overture-GoTo? Yeah, I think it was and there was a legal hassle and a billion dollar sentiment to make the GOOG gleam like a sprightly young Googzilla.
The actions that cement the frenzy in the House of Google is the steady flow of “it’s coming,” “yes, it’s a demo,” and “okay, we bought an outfit that Sam Bankman Fried found interesting.” The problem is that “to be” does not close the gap with the ChatGPT riding its hyper-drive electronic motorcycle on Google’s private motorways.
Several observations:
- Will the OpenAI and ChatGPT thing help Microsoft address the security of its existing software and services? When?
- Will Microsoft milk the marketing buzz and return to business as usual: Killing printers, interfaces that baffle, and features that disrupt one’s activity on a Windows enabled system?
- Will Microsoft have an answer to those who would claim that smart software violates fair use of intellectual property?
- Will Microsoft be able to handle bias and outputs which lead to interesting but harmful outcomes like a student getting expelled after mommy and daddy paid $135,000 for tuition?
But for now. The payoff for Microsoft is the thrill of watching Googzilla squirm. And the “all” word? That’s just an illustration of the imprecision of Microsoftie speech.
Stephen E Arnold, February 9, 2023
Google and OpenAI: The Big Dust Up
February 8, 2023
Let’s go back to high school English class and a very demanding spinster named Miss Drake. Her test question was, “Who wrote this line?”
O heaven! that one might read the book of fate, and see the revolution of the times. (Henry IV, Part 2 [~ 1597] Act 3)
Do you remember? I do. The answer is the Bard of Avon, and he allegedly spun the phrase from his brain cells or he ripped it off from another playwright. Yeah, the plagiarism thing has not been resolved, and it is unclear if and with whom the Bard sucked in content and output a money-making play. Was the real Bard a primordial version of a creator on YouTube? Sure, why not draw that connection?
Now back to the quote. I like the idea of “the revolution of the times.”
The idea is that some wizard like Sundar or Prabhakar can check out the “Book of Fate” which may or may be among the works in the Google data warehouse and see the future. Just the Palantir seeing stone which works so darned well as those SPAC bets attest. Perhaps that’s what happened when Google declared a Code Red? Fear and a big bet that the GOOG can de-momentum ChatGPT.
When did OpenAI become a thing? I would suggest that it was in 2015 if one believes Sillycon Valley history. The next highlight of what was something of note took place in 2019 but possibly earlier when Microsoft plopped some Azure cycles on the OpenAI conference table. Two years later we get this:
Google at Code Red over ChatGPT As Teams Reassigned to Work on Competing AI Products
Almost coincident with Google’s realizing that people were genuinely excited about ChatGPT, Google realized that Prabhakar’s hair had caught on fire. Sundar, the Sillycon Valley manager par excellence called back the original Relevance Revolutionaries (Messrs. Brin and Page) after Microsoft made it evident to the well-fed at Davos that Softies were good at marketing. Maybe Microsoft fell short of the “Start Me Up” for the outstandingly “good enough” Windows 95, but the Chat GPT deal is notable. To make sure the GOOG got the message that Microsoft was surfing on the waves created by ChatGPT, another bundle of billions were allocated to OpenAI and ChatGPT. The time was January 2023, and it was clear that millions of norms interested in Microsoft’s use of ChatGPT in those “good enough” engineering marvels, Bing.com and the Google infused Edge browser.
Where are we on Wednesday, February 8, 2023? How about this as a marker:
Google said Bard would be widely available to the public in the next few weeks. Source: MSN.com
Yep, the killer words are right there—”would be.” Not here, the conditional future, just a limited test. Not integrated into heaven knows how many apps like OpenAI. Not here like the collection of links generated by Matt Shumer. Not here like the YouTube videos explaining how to build an app from scratch with ChatGPT. Nope. Google is into the “to be” and “demo” mode.
Let’s do simple math on Google’s situational awareness:
- 2015 OpenAI and Elon Musk are visible
- 2019 Microsoft provides some “money” which may be a round trip to pay for Azure cycles
- 2022 (November) ChatGPT gets a million users in five days
- 2022 (December) Google feels the heat from burning hair and screeches “Code Red”
- 2023 (January) Davos talks about ChatGPT, not just the new world order, power, money, and where to eat dinner
- 2023 (February) Google says, “Out version of ChatGPT is coming… soon. Really, very soon.” But for now it’s a limited demo. And Microsoft? The ChatGPT thing turned up when one of my team ran a query on Tuesday, February 7, 2023. Yep, ready or not there it was.
Several observations:
- The Sundar and Prabhakar duo missed the boat on the impact of ChatGPT for search. Err, you are supposed to be the search wizards, and you are standing on the platform waiting for the next train to arrive?
- The uptake of ChatGPT may be a reaction against the Google search system, not a reaction to the inherent greatness of ChatGPT. A certain search company with a 90 percent share and a genuine disdain to relevant responses to users’ queries may have boosted the interest in ChatGPT. If I am correct, this is an unintended consequence of being Googley.
- The Microsoft marketing move is an outstanding [a] bit of luck, [b] a super star tactic that warrants a Grammy (oh, wait, the Grammies suck), or [c] a way to breathe new life into products which suffer from featuritis and a lack of differentiation.
Net net: Sundar and Prabhakar are destined for Harvard case study glory. Also, the dynamic duo may pull a marshmallow from the fire, but will it make a great S’more? Does sophomoric rhyme with Anthropic. And Code Red? The revolution of the times might be here as the Bard wrote or obtained from a fellow playwright.
Stephen E Arnold, February 8, 2023
An Interesting View of AI Deployment
February 8, 2023
Let me summarize the 5,000 word essay “Let’s Speed Up AI”: Go faster. The idea is that one has to accelerate that which is already accelerating. Imagine a downhill skier with a jet pack. Once momentum is gained, fire the jet back pack. The author believes that wide, rapid uptake of AI methods will result in the types of applications and controls. I certainly am not keen on having Terminator kick in my door, but I will reserve judgment on the wisdom of this “technology will work out its kinks.” A French guy pointed out that technology has unanticipated consequences. Today it is not necessary to get one’s priestly robes in a twist.
The write up states:
This idea that we can imagine every problem before it happens is a bizarre byproduct of a big drop in our risk tolerance as a society.
Are countries with access to AI technology risk averse? That’s a question I am not able to answer. I am not sure anyone can. Perhaps it is a job for the to-be systems from Microsoft Bing or the Code Red Google?
In reference to the suggestion that AI has to slow down, the write up says:
The basic premise of the slow-down article is that AI doom is inevitable. We’ve got to slow it down or stop the research now to avoid the disaster! It’s not that AI might go crazy and kill us all, it’s that it will kill us all!
If you are fans of the go-fast approach to technology, you may find the “Let’s Speed Up AI” argument on the money. I think that some pharma execs like the idea of go-fast.
Is it important that Google with its oodles of smart software and thousands of wizards [a] overlooked the potential of ChatGPT, [b] noted the demand and did not know how to frame Googzilla’s response, or [c] has too many executives like Sundar and Prabhakar who want to go slow or not go anywhere that would undermine their compensation?
Google’s new posture is go fast. I wonder if the firm has weighed objectively the risks of rolling out a knee jerk response to what is the marketing home run of the last six months.
My hunch is that the battle is not among smart software vendors; I think we are engaged in a marketing tussle. Technology is like the illustrations of new cars in a 1950s’ Saturday Evening Post. The cars were not that good. But the messages and images were outstanding.
Accelerating the accelerating sounds good. Should we ask the downhill skier with the jet pack?
Stephen E Arnold, February 8, 2023
Google: So Clever, So So Clever
February 6, 2023
I read a good summary of the US and state governments’ allegations about the behavior of the Google ad machine. I recommend “How Google Manipulated Digital Ad Prices and Hurt Publishers, Per DOJ.” The write up provides some useful insight into how the Google management environment has created a culture of being really cute, possibly really clever. The methods employed reminded me of a group of high school science club members pranking the hapless administration of a secondary school. Fun and being able to be smarter than everyone else is the name of the game.
Let me cite one example from the write up because it is short, to the point, and leaves little room for a statement like, “Senator, I did not know how the system’s components worked. I will provide the information you need. Again, I am sorry.” Does that line sound familiar? I left out the “Senator, thank you for the question” but otherwise the sentiment seems in tune with the song some companies sing to semi-aware elected officials.
Google Ads allegedly submitted two bid prices, unbeknownst to advertisers and publishers, effectively controlling the winning bids and the price floors. To entrench its market power even further, the suit argues Google started manipulating ad prices under a different method, which it dubbed “Bernanke.” Starting in 2013, according to the suit, Google Ads would submit bid prices to AdX above the amount advertisers had budgeted, in order to win high-value impressions for a group of publishers — the ones most likely to switch ad tech platforms. This insight could only be obtained by leveraging data in Google’s own publisher ad server. Once AdX cleared the bids, Google Ads would offset the losses by charging higher fees to other publishers less likely to switch ad tech providers. This scheme allegedly helped Google lock in key publishers away from other ad exchanges and ad buying tools, all while maintaining its profits at the expense of other smaller publishers.
Once of the best jobs I had in my life was my stint at the Courier-Journal & Louisville Times Co. That newspaper, like many others, has been unable to cope with the digital revolution. Outfits like Google and their clever methods may have hastened the financial precipice on which many publishers teeter.
My concern is that this particular method — just one of many I assume — has been grinding out cash for the Google for about a decade. Now there is some action, but I think the far more important challenge Google faces will be the active consumer uptake of newer options. These may prove to be familiar with Clever Avenue.
I hope these AI-informed travelers take the road called Ethical Behavior Boulevard.
Stephen E Arnold, February 6, 2023
-
- Subscribe to Beyond Search
Feature archive
News archive
-
