Open AI and Its Alignment Pipeline

July 12, 2023

Vea4_thumb_thumb_thumb_thumb_thumb_t[1]Note: This essay is the work of a real and still-alive dinobaby. No smart software involved, just a dumb humanoid.

Yep, alignment pipeline. No, I have zero clue what that means. I came across this felicitous phrase in “OpenAI Co-Founder Warns Superintelligent AI Must Be Controlled to Prevent Possible Human Extinction.” The “real news” story focuses on the PR push for Sam AI-Man’s OpenAI outfit. The idea for the story strikes me as a PR confection, but I am a dinobaby. Dinobabies can be skeptical.

7 6 bullshit artist

An OpenAI professional explains to some of his friends that smart software may lead to human extinction. Maybe some dogs and cockroaches will survive. He points out that his company may save the world with an alignment pipeline. The crowd seems to be getting riled up. Someone says, “What’s an alignment pipeline.” A happy honk from the ArnoldIT logo to the ever-creative MidJourney system. (Will it be destroyed too?)

The write up reports a quote from one of Sam AI-Man’s colleagues; to wit:

“Superintelligence will be the most impactful technology humanity has ever invented, and could help us solve many of the world’s most important problems. But the vast power of superintelligence could also be very dangerous, and could lead to the disempowerment of humanity or even human extinction,” Ilya Sutskever and head of alignment Jan Leike wrote in a Tuesday blog post, saying they believe such advancements could arrive as soon as this decade.

There you go. Global warming, the threat of nuclear discharges in Japan and Ukraine, post-Covid hangover, and human extinction. Okay

What’s interesting to this dinobaby is that OpenAI made a decision to make the cloud service available. OpenAI hooked up with the thoughtful, kind, and humane Microsoft. OpenAI forced the somewhat lethargic Googzilla to shift into gear and respond.

The Murdoch article presents another OpenAI wizard output:

“Currently, we don’t have a solution for steering or controlling a potentially superintelligent AI, and preventing it from going rogue. Our current techniques for aligning AI, such as reinforcement learning from human feedback, rely on humans’ ability to supervise AI. But humans won’t be able to reliably supervise AI systems much smarter than us and so our current alignment techniques will not scale to superintelligence,” they wrote. “We need new scientific and technical breakthroughs.”

This type of jibber jabber is fascinating. I wonder why the OpenAI folks did not do a bit of that “what if” thinking before making the service available. Yeah, woulda, shoulda, coulda. It sounds to me like a driver saying to a police officer, “I didn’t mean to run over Grandma Wilson.”

How does that sound to the grand children, Grandma’s insurance company, and the judge?

Sounds good, but someone ran over Grandma Wilson, right, Mr. OpenAI wizards? Answer the question, please.

The OpenAI geniuses have an answer, and I quote:

To solve these problems, within a period of four years, they said they’re leading a new team and dedicating 20% of the compute power secured to date to this effort. “While this is an incredibly ambitious goal and we’re not guaranteed to succeed, we are optimistic that a focused, concerted effort can solve this problem,” they said.

Now the capstone:

Its goal is to devise a roughly human-level automated alignment researcher, using vast amounts of compute to scale it and “iteratively align superintelligence.” In order to do so, OpenAI will develop a scalable training method, validate the resulting model and then stress test its alignment pipeline.

Yes, the alignment pipeline. What a crock of high school science club yip yap. Par for the course today. Nice thinking, PR people. One final thought: Grandma is dead. CYA words may not impress some people. To a high school science club type, the logic and the committee make perfect sense. Good work, Mr. AI-Men.

Stephen E Arnold, July 12, 2023

Amazon Is Winning the Product Search Derby… for Now

July 12, 2023

Vea4_thumb_thumb_thumb_thumb_thumb_t[1]Note: This essay is the work of a real and still-alive dinobaby. No smart software involved, just a dumb humanoid.

Google cannot be happy about these numbers. We learn from a piece at Search Engine Land that now “50% of Product Searches Start on Amazon.” That is even worse for the competition than previously predicted. In fact, Google’s share of this market has slipped to less than a third at 31.5%. What’s Google’s solution to this click loss? Higher ad pricing? Or maybe an even higher ad-to-real content ratio?

7 9 search race

The search racers are struggling to win traffic related to products. What has Amazon accomplished? Has Google’s vehicle lost power? What about Microsoft, a company whose engine is Bing-ing?

We also learn just 14% of respondents start their searches at retail or brand websites, while social media and review sites each capture a measly 2%. But that could change as Generation Z continues to age into independent shoppers. That group is the most likely to launch searches from social media. They are also most inclined to check online reviews. Reviews with photos are especially influential. Writer Danny Goodwin cites a recent Pew survey as he writes:

“Reviews and ratings can make or break a sale more than any other factor, including product price, free shipping, free returns and exchanges, and more. Overall, 77% of respondents said they specifically seek out websites with reviews – and this number was even higher for Gen Z (87%) and millennials (81%). Ratings without accompanying reviews are considered untrustworthy by 56% of survey respondents. Where people read reviews and ratings:

  • Amazon: 94%
  • Retail websites (e.g., Target, Wal-Mart): 91%
  • Search engines: 70%
  • Brand websites (the brand that manufactures the product: 68%
  • Independent review sites: 40%

User-generated photos and videos gain value. Sixty percent of consumers looked at user-generated images or videos when learning about new products.

  • 77% of respondents said they trust customer photos and videos.
  • 53% said user-generated photos and videos from previous customers impacted their decision whether to purchase a product.”

So there you have it—if you have a product to market online, best encourage reviews. With pics, or it didn’t happen. Videos are a significant marketing factor. What happens if Zuck’s Threads pushes into product search, effectively linking text promotions with Instagram? And the Google? Let’s ask Bard?

Cynthia Murrell, July 12, 2023

On Twitter a Personal Endorsement Has Value

July 11, 2023

Vea4_thumb_thumb_thumb_thumb_thumb_t[1]Note: This essay is the work of a real and still-alive dinobaby. No smart software involved, just a dumb humanoid.

The high school science club managers are engaged in a somewhat amusing dust up. First, there was a challenge to a physical fight, a modern joust in which two wizards would ride their egos into glory in Las Vegas, a physical metaphor for modern America. Then the two captains of industry would battle in court because … you know… you cannot hire people another company fired. Yesterday, real journalists crowed from many low rise apartment roof tops that a new social media service was growing allegedly at the expense of another social media company. The numbers prove that one company is better at providing a platform to erode cultural values than another. Victory!

7 11 truck scene

Twitter… endorsed by those who know. Thanks, MidJourney, you output an image in spite of your inappropriate content filter. Good work.

Now I learn that one social media outfit is the bestie of an interesting organization. I think that organization has been known to cast aspersions on the United States. The phrase “the great Satan” sticks in my mind, but I am easily confused. I want to turn to a real news outfit which itself is the subject of some financial minds — Vice Motherboard.

The article title makes the point: “Taliban Endorses Twitter over Threads.” Now that is quite an accolade. The Facebook Zucker service, according to the article, is “intolerant.” Okay. Is the Taliban associated with lenient and tolerant behavior? I don’t know but I recall some anecdotes about being careful about what to wear when pow-wowing with the Taliban. Maybe that’s incorrect.

The write up adds:

Anas Haqqani, a Taliban thought-leader with family connections to leadership, has officially endorsed Twitter over Facebook-owned competitor Threads. “Twitter has two important advantages over other social media platforms,” Haqqani said in an English post on Twitter. “The first privilege is the freedom of speech. The second privilege is the public nature & credibility of Twitter. Twitter doesn’t have an intolerant policy like Meta. Other platforms cannot replace it.”

What group will endorse Threads directly and the Zuck implicitly? No, I don’t have any suggestions to offer. Why? This adolescent behavior can manifest itself in quite dramatic ways. As a dinobaby, I am not into drama. I am definitely interested in how those in adult bodies act out their adolescent thought processes. Thumbs up for Mr. Musk. Rocket thrusters, Teslas, and the Taliban. That’s the guts of an impressive LinkedIn résumé.

Stephen E Arnold, July 11, 2023

Adolescent Technology Mavens: From the Cage to the Court House

July 11, 2023

Vea4_thumb_thumb_thumb_thumb_thumb_t[1]Note: This essay is the work of a real and still-alive dinobaby. No smart software involved, just a dumb humanoid.

Ladieees and gentlemennnnn, in this corner we have the King of Space and EVs. Weighing 187.3 pounds, the Musker brings a devastating attitude and a known world class skill in naming things. With a record of three and one, his only loss was a self-inflicted KO fighting a large blue bird. Annnnd in this corner, we have the regulator’s favorite wizard, Mark the Eloquent. Weighing in at 155.7 pounds, the Zuckster has a record of 3 and 3. His losses to Cambridge Analytica, the frightening Andrea Jelinek, chair of the European Data Protection Board, and his neighbor in Hawaii who won’t sell land to the social whirlwind.

Where are these young-at-heart wizards fighting? In Las Vegas for a big pile of money? Nope. These estimable wizards will duke it out in the court house. “Scared Musk Sends Legal Threat to Meta after Threads Lures 30 Million on Launch Day” states as fresh-from-the-playground news:

Musk supplemented his tweet [https://twitter.com/elonmusk/status/1676770522200252417] with a legal threat against Meta that echoed despair and fear in the face of his potent adversary. The lawsuit alleges Meta of enticing Twitter’s former employees — many of whom Musk dismissed without honoring severance promises — to contribute to Threads, a move that Twitter asserts infringes upon its intellectual property rights.

One big time journalist took issue with my describing the senior managers of certain high technology firms as practicing “high school science club management methods.” I wish to suggest that rumored cage fight and the possible legal dust up illustrates the thought processes of high school science club members. Yeah, go all in with those 16-year-old decision processes.

The threads are indeed tangled.

Stephen E Arnold, July 11, 2023

Scinapse Is A Free Academic-Centric Database

July 11, 2023

Vea4_thumb_thumb_thumb_thumb_thumb_t[1]Note: This essay is the work of a real and still-alive dinobaby. No smart software involved, just a dumb humanoid.

Quality academic worthy databases are difficult to locate outside of libraries and schools. Google Scholar attempted to qualify as an alternative to paywalled databases, but it returns repetitive and inaccurate results. Thanks to AI algorithms, free databases improved, such as Scinapse.

Scinapse is designed by Pluto and it is advertised as the “researcher’s favorite search engine. Scinapse delivers accurate and updated research materials in each search. Many free databases pull their results from old citations and fail to include recent publications. Pluto promises Scinapse delivers high-performing results due to its original algorithm optimized for research.

The algorithm returns research materials based on when it was published, how many times it was citied, and how impactful a paper was in notable journals. Scinapse consistently delivers results that are better than Google Scholar. Each search item includes a complete citation for quick reference. The customized filters offer the typical ways to narrow or broaden results, including journal, field of study, conference, author, publication year, and more.

People can also create an account to organize their research in reading lists, share with other scholars, or export as a citation list. Perhaps the most innovative feature is the paper recommendations where Scinapse sends paper citations that align with research. Scinapse aggregates over 48,000 journals. There are users in 196 countries and 1,130 reputable affiliations. Scinapse’s data sources include Microsoft Research, PubMed, Semantic Scholar, and Springer Nature.

Whitney Grace, July 11, 2023

In the Midst of Info Chaos, a Path Identified and Explained

July 10, 2023

Vea4_thumb_thumb_thumb_thumb_thumb_t[1]Note: This essay is the work of a real and still-alive dinobaby. No smart software involved, just a dumb humanoid.

The Thread – Twitter spat in the midst of BlueSky and Mastodon mark a modest change in having one place to go for current information. How does one maintain awareness with high school taunts awing, Mastodon explaining how easy it is to use, and BlueSky doing its deep gaze thing?

One answer and a quite good one at that appears in “RSS for Post-Twitter News and Web Monitoring.” The author knows quite a bit about finding information, and she also has the wisdom to address me as “dinobaby.” I know a GenZ when I get an email that begins, “Hey, there.” Trust me. That salutation does not work as the author expects.

In the cited article, you will get useful information about newsfeeds, screenshots, and practical advice. Here’s an example of what’s in the excellent how to:

If you want to check a site for RSS feeds and you think it might be a WordPress site, just add /feed/ to the end of the domain name. You might get a 404 error, but you also might get a page full of information!

There are more tips. Just navigate to Research Buzz, and learn.

This dinobaby awards one swish of its tail to Tara Calishain. Swish.

Stephen E Arnold, July 10, 2023

A Lesson in Negotiation: A Scholarly Analysis of the Musk-Zuck Interaction

July 10, 2023

Vea4_thumb_thumb_thumb_thumb_thumb_t[1]Note: This essay is the work of a real and still-alive dinobaby. No smart software involved, just a dumb humanoid.

Zuck Is a Cuck: Elon Musk Ramps Up His Attacks on Mark Zuckerberg With Shocking Tweet” provides an example of mature decision making, eloquent rhetoric, and the thrill of the high school insult. Maybe, it is a grade-school thrill, similar to someone pointing at overweight me with thick glasses and a book to read just for fun. I can hear the echoes of these memorable words, “Look at smarty pants. Yah yah yah.” I loved every minute of these insults.

7 10 teens fight

“What did you call me? You keep your mouth shut or my friends and I will post on both Threads and Twitter that you do drugs and steal to buy junk.” Yes, the intellectual discourse of those in the prime of adolescence. And what’s the rejoinder, “Yeah, well, I will post those pix you sent me and email them to your loser mom. What do you think about that, you, you [censored]?”

The cited article from Mediaite (which I don’t know how to pronounce) reports:

Threads drew tens of millions of users since its launch three days ago, so the competition between Musk and Zuckerberg is being waged on social, legal, and perhaps even physical fronts with talk of a cage match fight between the two. Despite the numerous setbacks Twitter has seen since Musk took it over, he has spent the weekend hyping up improvements to the platform while taking shots at Zuckerberg.

What business school teaching moment is this? [a] Civil discourse triumphs, [b] Friendly competition is a net positive, [c] Ad hominem arguments are an exceptional argumentative tool, [d] Emotional intelligence is a powerful opportunity magnet.

What? Why no [e] All of the above?

Note for those who don’t like my characterization of Silicon Valley luminaries’ manifestation of “the high school science club management method. Isn’t it time to accept HS-SC-MM as the one “true way” to riches, respect, and power?

Stephen E Arnold, July 10, 2023

TikTok Interface: Ignoring the Big Questions

July 10, 2023

Vea4_thumb_thumb_thumb_thumb_thumb_t[1]Note: This essay is the work of a real and still-alive dinobaby. No smart software involved, just a dumb humanoid.

I read “TikTok Is Confusing by Design.” That’s correct. But the write up does not focus on the big questions. However, the article tiptoes up to the $64 question and then goes for a mocha latte. Very modern.

7 6 ignore red light

A number of articles ignore flashing red lights. William James called this “a certain blindness.” Thanks, MidJourney to a wonderful illustration crafted from who knows what.

Note these snippets from the essay:

  • a controlled experience that’s optimized to know or decide what we want and then deliver it to us.
  • You don’t get to choose from a list of related content, nor is there any real order to whatever you’ll get.
  • It’s a comfortable space to be in when you don’t have to make choices.
  • TikTok’s approach has become the new standard. Part of that standard is aggressively pushing content at you that the app has decided you want to see.

So what are the big questions? The article shoves them to the end of the essay. Will people persist and ponder them? Don’t big questions warrant a more compelling presentation?

Here’s a big question:

“Who gets to control what you are seeing of reality?”

The answer is obvious in the case of TikTok: Entities in some way linked to the Chinese government.

And what about online services working overtime to duplicate the TikTok model? Who is in control of the content, its context, and its concepts?

The answer is, “An outfit that will have unprecedented amount of influence over users’ thoughts and actions.” If those users — digital addicts, perhaps — are not able to recognize manipulation or simply choose to say, “Hey, no big deal”, TikTok-type content systems will be driving folks down the Information Highway. Riders may have no choice. Riders may have to pay to driven around. Riders may not be in control of their behaviors, ideas, and time.

I like the idea of TikTok as an interface. I don’t like touching on big questions and then sidestepping them.

Net net: I won’t pay for access to Vox.

Stephen E Arnold, July 10, 2023

Google and AMP: Good Enough

July 10, 2023

Vea4_thumb_thumb_thumb_thumb_thumb_t[1]Note: This essay is the work of a real and still-alive dinobaby. No smart software involved, just a dumb humanoid.

Due to the rise of mobile devices circa the 2010s, the Internet was slammed with slow-loading Web-sites. In 2015, Google told publishers it had a solution dubbed “Accelerated Mobile Pages” (AMP). Everyone bought into AMP but it soon proved to be more like a “Speed Trap” says The Verge.

AMP worked well at first but it was hard to use advertising tools that were not from Google. Google’s plan to make the Internet great again backfired. Seventeen state attorneys filed a lawsuit with AMP as a key topic against Google in 2020. The lawsuit alleges Google purposefully designed AMP to prevent publishers from using alternative ad tools. The US Justice Department filed an antitrust lawsuit in January 2023, claiming Google is attempting to control more of the Internet.

79 googzilla

A creature named Googzilla chats with a well-known publisher about a business relationship. Googzilla is definitely impressed with the publisher’s assertion that quality news can generate traffic and revenue without a certain Web search company’s help. Does the publisher trust Googzilla? Sure, the publisher says, “We just have lunch and chat. No problem.” 

Google promised that AMP would drive more traffic to publishers’ Web sites and it would fix the loading speed lag. Google was the only big tech company that offered a viable solution to the growing demand mobile devices created, so everyone was forced to adopt AMP. Google did not care as long as it was the only player in the game:

“As long as anyone played the game, everybody had to. ‘Google’s strategy is always to create prisoner’s dilemmas that it controls — to create a system such that if only one person defects, then they win,’ a former media executive says. As long as anyone was willing to use AMP and get into that carousel, everyone else had to do the same or risk being left out.”

Google promised AMP would be open source but Google flip-flopped on that decision whenever it suited the company. Non-Google developers “fixed” AMP by working through its locked down structure so it could support other tools. Because of their efforts AMP got better and is now a decent tool. Google, however, trundles along. Perhaps Google is just misunderstood.

Whitney Grace, July 10, 2023

Whom Does One Trust? Surprise!

July 7, 2023

Vea4_thumb_thumb_thumb_thumb_thumb_t[1]Note: This essay is the work of a real and still-alive dinobaby. No smart software involved, just a dumb humanoid.

Whom does one trust? The answer — according to the estimable New York Post — is young people. Believe it or not. Just be sure to exclude dinobabies and millennials, of course.

Millennials Are the Biggest Liars of All Generations, Survey Reveals”:

A new survey found that of all generations, those born between 1981 and 1996 are the biggest culprits of lying in the workplace and on social media.

Am I convinced that the survey is spot on? Nah. Am I confident that millennials are the biggest liars when the cohort is considered? Nah.

7 6 liar

MidJourney generated this illustration of an angry sales manager confronting a worker. The employee reported sales as closed when they were pending. Who would do this? A dinobaby, a millennial, or a regular sales professional?

Am I entertained by the idea that dinobabies are not the most prone to prevarication and mendacity? Yes.

Consider this statement:

The findings showed that millennials were the worst offenders, with 13% copping to being dishonest at least once a day.

How many times do dinobabies eject a falsehood?

By contrast, only 2% of baby boomers, those born between 1946 and 1964, fibbed once per day.

One must be aware that GenXers just five percent engage in “daily deception.”

Where do people take liberties with the truth? Résumés (hello, LinkedIn) and social media. Imagine that! Money and companionship.

Who lies the most? Yep, 26 percent of males lie once a day. Twenty-three percent of females emit deceptive statements once a day. No other genders were considered in the write up, which is an important oversight in my opinion.

And who ran the survey? An outfit named PlayStar. Yes! I wonder if the survey tool was a Survey Monkey-like system.

Stephen E Arnold, July 7, 2023

« Previous PageNext Page »

  • Archives

  • Recent Posts

  • Meta