<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Fred Abler on Medium]]></title>
        <description><![CDATA[Stories by Fred Abler on Medium]]></description>
        <link>https://medium.com/@fredabler?source=rss-b5518b5cf762------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*_y7BD1XINXaLZhCB.</url>
            <title>Stories by Fred Abler on Medium</title>
            <link>https://medium.com/@fredabler?source=rss-b5518b5cf762------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Wed, 18 Nov 2026 14:44:57 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@fredabler/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[AI’s Killer App is illiteracy]]></title>
            <link>https://medium.com/@fredabler/ais-killer-app-is-illiteracy-2f8d512dbad2?source=rss-b5518b5cf762------2</link>
            <guid isPermaLink="false">https://medium.com/p/2f8d512dbad2</guid>
            <dc:creator><![CDATA[Fred Abler]]></dc:creator>
            <pubDate>Sun, 01 Mar 2026 22:29:44 GMT</pubDate>
            <atom:updated>2026-03-03T19:36:48.945Z</atom:updated>
            <content:encoded><![CDATA[<p>I’m sorry Professor Nguyen ( I owe you $5 in chocolate math)</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*a5VMBFn4Vl8Tvc2cJZo3LA.png" /></figure><h4><strong><em>Why should you read anything.. if AI has already read everything?</em></strong></h4><p>I realize this question sounds like some smart-ass college student’s lazy logic. But the question seriously confounded me this morning, after reading Derek Thompson’s excellent post on <a href="https://www.derekthompson.org/p/how-metrics-make-us-miserable?utm_source=post-email-title&amp;publication_id=2880588&amp;post_id=189384015&amp;utm_campaign=email-post-title&amp;isFreemail=true&amp;r=4ysyj&amp;triedRedirect=true&amp;utm_medium=email"><strong>How Metrics Make Us Miserable.</strong></a></p><p>In Thompson’s post, he interviews <strong><em>C. Thi Nguyen</em></strong>, who is the author of the newly released book <strong><em>The Score: How to Stop Playing Somebody Else’s Game</em>. </strong>Dr. Nguyen is the University of Utah associate philosophy professor — a chair specifically designed to connect with the computer science and game design programs — through research collaborations and co-teaching.</p><p>The interview was good enough that I suddenly found myself on Amazon, considering buying Nguyen’s latest book. But it‘s only available in hard cover? Meh! I no longer really care about building a trophy bookshelf. And full disclosure, playing someone else’s game — has never been my problem.</p><p>So $30 seemed like a lot of money to drop on a dead-tree edition. I am also rarely able to actually finish reading a book. Back in my day, the world was only in black and white, and there was no such thing as ADHD. So, does everyone else actually ‘finish’ reading books? Or, is it adult-onset ADHD?</p><p>In any case, I really hate paying good money to make myself feel guilty. It does not feel empowering. So did I really didn’t want to spend $30 — on a book I’ll never finish? And that’s when my AI ‘lazy logic’ hit me — like a ton of books.</p><p><strong>“Why should I read anything.. AI has already read everything?</strong></p><p>I mused at my own newfound internalization of AI. Indeed! I congratulated myself as I smugly prompted Claude.ai to give me the top 3 insights from Nguyen’s other <strong><em>Agency as Art</em></strong> book, which it did almost instantly.</p><p><strong>Ironic Editor’s Note: I had included the Claude results here, but my editor (Google’s Gemini AI) told me ‘no-one ever reads these long pull quotes’ and that I was likely ‘stalling’ my readers. See, even AI knows that humans don’t read…</strong></p><p>The delight at saving myself $30, and at capturing a few years of Dr. Nguyen’s life-work in seconds — was immediate. Alas, the book was not about the kind of <strong>AI </strong>Agents I am interested in. It’s about how games give players a theatrical ‘agent’ like experience.</p><p>But almost as immediate was my quick pang of academic guilt. I had just deprived the good professor of any real income for his hard work. And if you think reading is difficult — <strong><em>writing</em></strong> anything (including books) is hella’ hard work.</p><p>My father was a professor who wrote a still widely used text book in geography. In the early 1970’s, his royalty was almost $1 per book sold. I only know this because, one day when I was 8 years old, we went to a grocery store, where he purchased 35 ‘Giant’ 8 oz Hershey’s chocolate bars.</p><p>He didn’t want to profit from making his own book required reading, so he “kicked back” his $1-per-book royalty to his university students - in chocolate.</p><p>Adjusting for inflation, I figure that I owe Professor Nguyen $4.49. That’s the price of the ‘Giant’ Hershy’s bar today. But thanks to shrinkflation the bar is now only 7.56 oz. So to be safe, let’s round up to $5.</p><p>It doesn’t sound like much money but that’s half the cost of a decent bottle of French Bordeaux wine from Trader Joe’s. Or, an <strong><em>excellent full bottle </em></strong>of Sharon Weeks’ CATTOO — a punk-rock zinfandel from Paso Robles, CA which is only $3.99 at the Grocery Outlet.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1022/1*CNt_KK441TDIqONgUfxqaQ.png" /></figure><p>No matter how you spend it, $5.00 US is still real money. Hard currency, red wine or chocolate that Professor Nguyen never gets. Not to mention all the associated revenue lost to: the university, the book’s publisher, the printer, and yes, even poor Amazon. (They’re so not-poor).</p><p><strong>Conclusion:</strong> Our “lazy logic” has some real economic blood on its hands. Before AI, only Cliff Notes were available to avoid the ‘cognitive load’ of long-reading. But now there is precious little that cannot be encapsulated, condensed, or easily refactored by GenAI.</p><p>We are currently worried about AI taking jobs from software engineers. It’s a vaild concern, but making software on-demand is not as easy as many AI doomers think. Sure you can do it, but you can also build your own car in the barn — provided you have 5 years on your hands, and no-life to sacrifice.</p><p>Saas-A-geddon? I don’t think so.</p><p>Likely we should be more worried that the “Killer App” of AI is <strong><em>illiteracy. </em></strong>What happens when people who actually make <strong><em>new knowledge — </em></strong>can no longer can get paid for their brain-work? Pretty certain it’s nothing good! And I’m not even factoring-in the deleterious impacts on our human culture, at large.</p><p>(C) 2025 FRED</p><p><strong>#ArtificialIntelligence, </strong>#FutureOfWork, <strong>#GenerativeAI, #</strong>Philosophy, <strong>#Authorship, #Amazon, #CATTOO, #TraderJoes, #GroceryOutlet, #UniversityofUtah</strong></p><p>P.S. The chocolate bar in the cover image is not to scale. It should be much larger. A standard notecard is 3&quot;x 5&quot;, But try as I might, I was unable to coax Nano Bannana to make the chocolate bar the correct size ( <strong>7.79 inches in length, 4.1 inches in width)</strong> It will be powerful when GenAI has state, and an inbuilt sense of relative scale, making such edits easy.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=2f8d512dbad2" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[TACO BELL’s ‘Human OS’ is fighting its Voice AI]]></title>
            <link>https://medium.com/@fredabler/taco-bells-human-os-is-defeating-its-voice-ai-e199eb373557?source=rss-b5518b5cf762------2</link>
            <guid isPermaLink="false">https://medium.com/p/e199eb373557</guid>
            <dc:creator><![CDATA[Fred Abler]]></dc:creator>
            <pubDate>Fri, 06 Feb 2026 17:24:03 GMT</pubDate>
            <atom:updated>2026-02-10T18:26:51.452Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*LAbvHeEDC8pbEcBmSRLbNg.jpeg" /></figure><p>Since I first wrote about Taco Bell ‘<a href="https://medium.com/@fredabler/crunchwrapping-ai-at-taco-bell-28c09791f8ab">Crunchwrapping AI</a>’ a year ago, only a few things have changed. Taco Bell’s drive-thru AI is now rebuilt using Omilia’s Voice AI, running on the “Byte by Yum” platform. This system <em>was</em> supposed to be a comprehensive ordering solution.</p><p>The intentions were sound (pun intended), but something unexpected is now happening: the ‘Human OS’ — the human operating system around the AI — is not cooperating.</p><p>Last night, I re-tested the AI with my patented ‘3 Supreme’ Test, where each taco is customized differently. This test consistently stumps the AI. But watching the screen as I ordered, the latest AI <em>finally</em> showed an object-model for exceptions and substitutions, neatly indented.</p><p>I thought the Omilia’s Voice AI might actually pass my Kobayashi Maru!</p><p>But alas, the on-screen customizations were wildly wrong. The AI seems to be using the WAG Algorithm (Wild Ass Guessing). I waited for a human meat puppet to step in, but they never did — so I surrendered to the chaos of random tacos. Still, I had no idea how weird things were about to get.</p><p>When I dissected my bag-bomb in the parking lot, the order I received was nowhere near TRUE to the AI’s already mangled order. NONE of the Tacos had Cheese! And how does ‘No Sour Cream’ for 1 Taco become — No Shredded Cheese for ALL Tacos? Or ‘Black Beans instead of meat’ on 1 Taco becomes… ‘Black Beans on 3 Tacos?</p><p>I concluded it’s an algo-glitch called ‘The Drunken 3 Musketeers’. “One for All, and All for none!”</p><p>The real Answer here is: it is <strong>The Human OS</strong>. At Taco Bell, the Human OS is the kitchen crew, now officially called <strong>“Food Champions.” </strong>(You simply can’t make this stuff up).</p><p>The Human OS is the problem. While voice AI <strong><em>may</em></strong> eventually handle custom orders correctly, there’s likely to be zero follow-thru in the kitchen. In fact, the more complex and custom your AI order is, the higher the probability the Food Champions will just <strong><em>‘freestyle’</em></strong> your order.</p><p>After all, they know you’re going to literally drive-away immediately — without checking first.</p><p>So Here is my logical model for the new (AI vs. Human OS) disruption:</p><p>As AI Order Accuracy ($A$) increases, the Menu Order Complexity ($C$) increases geometrically. This can be modeled as an exponential relationship where the complexity grows rapidly as accuracy improves:</p><p>$$C = k_1 \cdot e^{m_1 \cdot A}$$</p><ul><li>$k_1$: A scaling constant (e.g., base complexity).</li><li>$m_1$: A positive constant that determines the rate of geometric increase.</li></ul><p><strong>Relationship 2: Menu Complexity and Human OS Factor</strong></p><p>As Menu Order Complexity ($C$) increases, the Human OS “Give a RIP” Factor ($R$) decreases exponentially. This is an inverse relationship showing a rapid loss of human motivation as the order becomes more complex:</p><p>$$R = k_2 \cdot e^{-m_2 \cdot C}$$</p><ul><li>$k_2$: A scaling constant (e.g., initial motivation level).</li><li>$m_2$: A positive constant that determines the rate of exponential decrease.</li></ul><p><strong>CONCLUSION:</strong> Taco Bell’s Voice AI isn’t great, but it’s slowly improving. It’s now good enough to see that even perfect voice <strong>AI</strong> Order-Taking will be (soundly) defeated by the Bell’s Gen-Z Food Champions. It’s called the ‘<strong>Human OS</strong>’ problem.</p><p>My brutal advice? YUM! BRANDS should immediately D’cease the Voice AI Experiment altogether, OR preferably — hire me for embarrassingly large consulting fees — so I can hire a test team that will relentlessly torture test the “AI + Food Champion” ORDER OUTCOMES — at all 500 locations.</p><p>Alas, until the Human OS problem is solved, there’s little profit in raising our collective expectations. And now that a ‘3 Supreme’ costs $15.00 US, we absolutely expect the Human OS to get our orders right.</p><p>PS: The VOLUME of the AI is STILL earsplittingly loud. It’s godawful and so abusive that I predict Taco Bell will be sued for hearing loss….</p><p>© 2025 FRED</p><p><a href="https://omilia.com/ai-meets-the-drive-thru-taco-bells-journey-to-automated-customer-service/">Omilia</a></p><p><a href="https://restauranttechnologynews.com/2025/08/how-taco-bells-ai-drive-thru-became-a-viral-sensation-for-the-wrong-reasons/">Restauranttechnologynews</a>.</p><p><a href="https://www.complex.com/pop-culture/a/treyalston/taco-bell-rethinks-ai-drive-thru-viral-videos">Complex</a>.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=e199eb373557" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Prediction Market Paradox:]]></title>
            <link>https://medium.com/@fredabler/the-prediction-market-paradox-lets-not-build-ai-on-a-foundation-of-organized-gambling-7789dd540e75?source=rss-b5518b5cf762------2</link>
            <guid isPermaLink="false">https://medium.com/p/7789dd540e75</guid>
            <dc:creator><![CDATA[Fred Abler]]></dc:creator>
            <pubDate>Sun, 28 Dec 2025 22:30:26 GMT</pubDate>
            <atom:updated>2026-01-25T17:03:28.889Z</atom:updated>
            <content:encoded><![CDATA[<h3>Building AI on a Foundation of Organized Gambling?</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*NClcJ0cVaRhUIDLB.jpg" /></figure><p>People have lost faith in institutions, yet they trust markets more than ever. And the irony of course, is that markets <em>are</em> institutions. This pithy paradox matters because it appears to be driving something even more bizarre — a rush to use nascent prediction markets (i.e., <a href="https://polymarket.com/">Polymarket</a>, <a href="https://kalshi.com/markets/kxnasdaq100u/nasdaq-abovebelow/kxnasdaq100u-25dec29h1000">Kalshi</a>, <a href="https://robinhood.com/us/en/prediction-markets/">RobinHood</a>, Coinbase, etc.) as the “human signal” for training AI systems.</p><p>This highly suspect idea — that analog prediction markets can ‘smooth’ AI’s algorithmic machine intelligence — is sophisticated misdirection. And it also reveals something profound about how poorly we are collectively processing (our perceived) institutional collapse.</p><h3>The Trust Asymmetry</h3><p>Markets do provide something traditional institutions can’t: immediate, tangible feedback. Your stock portfolio updates by the second. Products either sell or sit on shelves. Businesses succeed or fail visibly. These strong signals create <strong><em>an illusion</em></strong> of accountability that government agencies, universities, and regulatory bodies — operating on longer timescales with murkier outcomes — simply can’t match.</p><p>But the perceived difference runs deeper than just ‘speed’ and ‘transparency’. Markets offer the <strong><em>feeling</em></strong> of participation. You <em>choose</em> which stock to buy, which product to purchase. Institutions feel imposed. You can’t opt out of the FDA or the California DMV. So this heady sense of <strong><em>agency</em></strong> matters enormously to trust-building, even if that agency is largely illusory. (Most retail investors underperform index funds, after all.)</p><p>There’s also the <strong><em>distributed authority effect</em></strong>. When the GreenBay Packers win, locals say “We won!”. But when the Packers lose, instead they say “They lost”. Similarly, when Apple’s stock drops, “the market” decided — so no single bureaucrat is to blame. This diffusion of responsibility feels safer to us than concentrated institutional power, despite the fact that large investors and corporations are all wielding enormous influence.. behind the scenes.</p><p>Perhaps most importantly of late, markets appear to <strong><em>punish</em></strong> incompetence through bankruptcy and failure. Bad companies go out of business. Failed institutions? Well, they persist indefinitely. Nobody “fires” the Department of Motor Vehicles for poor customer service.</p><p>Yet the profound irony? Markets only function because we have robust institutions — <em>i.e.,</em> property rights, contract enforcement, regulatory frameworks, central banking, legal systems. The “free market” is itself a massive institutional achievement. But we’ve conveniently stopped seeing it that way.</p><h3>Enter Prediction Markets: Betting as “Risk Mitigation”</h3><p>Our institutional distrust has fueled the ongoing explosion of prediction markets like PolyMarket. The pitch is seductive: bet on outcomes AND mitigate risk simultaneously. But unless you’re a wheat farmer hedging your crop yield, this framing is logically absurd.</p><p>You weren’t exposed to “Trump winning the election” as a financial risk until you created that exposure.. <strong><em>by betting on it</em></strong>. The only “risk” you’re actually mitigating is the psychological discomfort that comes from <strong>uncertainty</strong>. And you’re paying for that illusory comfort by literally gambling.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*BgwuHU1OEsVg0vPwmoCJTA.png" /><figcaption>Polymarket is the leading marketplace for betting on almost any ‘Outcome’.</figcaption></figure><p>Why do people fall for this gambit ?</p><p><strong>The control premium.</strong> In an uncertain world, prediction markets offer the feeling of actually <strong><em>doing something </em></strong>about it. You can’t control whether an event happens, but you can control whether you profit from it. Hey, that’s something, right?</p><p><strong>The <em>information asymmetry fantasy</em>.</strong> Everyone thinks they have special insight others lack. Betting feels like capitalizing on your own unique knowledge, not gambling. So betting is not just control, but self-actualization.</p><p><strong><em>Hedonic framing</em>.</strong> Calling it “risk mitigation” or “hedging” sounds so much more sophisticated and prudent. Whereas calling it “gambling on political outcomes” just sounds degenerate. Same activity, different frame and different name.</p><p>The <strong><em>meta-irony.</em></strong> Prediction markets only work because most participants <em>aren’t</em> rationally hedging real exposure. They’re expressing opinions, gambling for entertainment, or virtue-signaling their beliefs. The real irony here is that “wisdom of crowds” requires a crowd of people making irrational bets.</p><h3>The Real Agenda: Prediction Markets as AI’s “Analog Taper”</h3><p>But gambling pathology/psychology isn’t the real story. The real impetus behind the prediction market boom, I believe, is far darker pattern and even more problematic. The real motivation is likely using betting on outcomes as some kind of <strong><em>epistemological</em></strong> input for AI systems.</p><p>The purported chain of reasoning likely goes like this:</p><ul><li>AI needs grounding in human preferences and knowledge</li><li>Human preferences are messy, subjective, contradictory</li><li>But markets convert preferences into clean price ‘signals’</li><li>Therefore: prediction markets = human knowledge API for AI</li></ul><p>It’s elegant, scalable, and also.. just completely wrong.</p><h3>Why This Is Epistemologically Bankrupt</h3><p><strong>Markets don’t reveal truth — they reveal bets about truth.</strong> When a prediction market says “Trump: 60% to win,” that’s not ground truth. It’s a price equilibrium between people with money, time, and motivated reasoning. So you’re feeding AI a measure of what bettors think, weighted by how much they’ll risk — all several layers removed from reality.</p><p>The garbage-in problem here is severe. Training or ‘factoring’ AI on prediction marketplace data means training it on:</p><ul><li>Whale manipulation (large bettors moving markets)</li><li>Recency bias (markets overreacting to latest news)</li><li>Liquidity constraints (thin markets producing bad signals)</li><li>Participant selection bias (who actually uses these platforms?)</li></ul><p>Then there’s also the circular reasoning trap. As AI gets better at forecasting, prediction markets will increasingly reflect what AI forecasts say, not independent human judgment. You end up with AI training on markets that are themselves shaped by AI predictions. An epistemological ouroboros eating its own tail. Or plainly…recursive garbage making.</p><h3>The Altman-Shaped Problem</h3><p>The AI capabilities hype machines, exemplified by reckless AI proponents like Sam Altman, have created this bizarre demand for an elaborate “alignment” infrastructure. For without alignment between obscene AI investments and real ROI outcomes, what else can rationalize this radical technology and Silicon Valley’s latest will-to-power?</p><p>Like AI itself, the prediction market hype represents:</p><p><strong><em>A solution looking for a problem.</em></strong> Do we really need human feedback at scale for pattern-matching systems? Or, are we pretending LLMs are AGI, and therefore require elaborate consensus mechanisms?</p><p><strong><em>Financialization as epistemology</em>.</strong> Converting every form of knowledge into a tradable asset (crypto, NFTs, fractional ownership, etc) is a bankrupt idea. Everything must have a market price, be on-chain, or it doesn’t exist? Truth itself gets a ticker symbol? Ah, if only it were that easy.</p><p><strong><em>Manufactured alignment theater</em>.</strong> Creating the appearance of grounding AI in human values while actually grounding it in… gambling behavior. It looks rigorous because there’s math and money involved. But it’s really not. But hey, at least it’s addictive, right?</p><h3>The Convergence of Hype Cycles</h3><p>What we’re witnessing is simply the super-collision of (at least) three distinct manias:</p><ol><li><strong>AI doomerism/boosterism</strong> creating demand for “alignment” solutions</li><li><strong>Crypto/prediction market evangelists</strong> seeking killer use cases. Anywhere they can</li><li><strong>Institutional distrust</strong> making “the market” seem like the only legitimate arbiter of truth</li></ol><p>Together these mega-trends produce a Frankenstein concept: AI that learns from prediction markets, which aggregate human wisdom, which in turn keeps AI aligned with human values.</p><p>But every link in that chain of thought contains either a fundamental category error or motivated reasoning. To quote Mel Brooks and Gene Wilder’s <strong><em>Young Frankenstein</em></strong>.. ‘ It’s… Abby Normal’.</p><h3>What Makes Markets “Real”?</h3><p>The implicit assumption is that financial stakes equal rational honesty. But we now have a century of Behavioral Economics proving this false. People are demonstrably terrible at probabilistic reasoning even when money is on the line.</p><p>What makes prediction markets more “real world” than structured surveys? Ethnographic research? Deliberative polling? Just because you can quantify something and trade it doesn’t make it more true. It just makes it more <em>liquid</em>. More <em>gamified</em>. More<em> fun</em>.</p><p>These aren’t genuine human inputs.. they’re a pale simulacra of human judgment that happens to be machine-readable. Pseudo-human signals optimized for algorithmic consumption, not for capturing what people actually think or value.</p><h3>The Darker Reading</h3><p>The really cynical interpretation: the people promoting this know it’s dubious, but they really ‘need this’- so they have to ignore the ample dangers:</p><ul><li>It creates new markets to trade on (monetization opportunity)</li><li>It positions AI companies as “responsibly seeking human input”</li><li>It delays reckoning with the fact that current AI can’t do what’s being promised</li><li>It manufactures plausible demand for blockchain infrastructure</li></ul><p>The prediction market boom isn’t about better forecasting or AI alignment. It’s about providing a psychologically satisfying way to engage with a world where traditional sense-making institutions have temporarily lost credibility — while creating new rent-extraction opportunities in the process.</p><h3>The Path Forward</h3><p>We can’t rebuild institutional trust by replacing human judgment with betting markets. We can’t align AI systems by training or ‘tapering’ them on the aggregated gambling behavior of traders. And we certainly can’t solve epistemological problems by throwing blockchain at them.</p><p>What we need instead:</p><ul><li>Honest reckoning with what current AI systems can and cannot do</li><li>Recognition that “the market” is not a neutral arbiter of truth</li><li>Investment in actually rebuilding legitimate institutional capacity</li><li>Skepticism toward solutions that conveniently serve multiple grift ecosystems simultaneously</li></ul><p>The ongoing prediction market surge reveals our collective desperation for <em>something</em> to trust. But building our AI future on organized or dis-organized gambling — and then feeding that to AI systems — isn’t wisdom. It’s just another layer of institutional (Venture Capital) failure, wrapped in the language of innovation.</p><p><em>Understanding why these dubious ideas gain traction is the first step toward demanding something more substantial.</em></p><p>© 2025 FRED.</p><p>#PredictionMarkets, #AIAlignment, #TechCriticism, #BehavioralEconomics, #InstitutionalTrust, #Polymarket, #Kalshi, #Robinhood</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=7789dd540e75" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[CrunchWrapping AI at Taco Bell]]></title>
            <link>https://medium.com/@fredabler/crunchwrapping-ai-at-taco-bell-28c09791f8ab?source=rss-b5518b5cf762------2</link>
            <guid isPermaLink="false">https://medium.com/p/28c09791f8ab</guid>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[voice-assistant]]></category>
            <category><![CDATA[fastfoodresturant]]></category>
            <category><![CDATA[hci]]></category>
            <category><![CDATA[ux]]></category>
            <dc:creator><![CDATA[Fred Abler]]></dc:creator>
            <pubDate>Mon, 27 Jan 2025 16:45:12 GMT</pubDate>
            <atom:updated>2026-01-30T16:20:10.845Z</atom:updated>
            <content:encoded><![CDATA[<p><strong>SoundHound </strong>AI provides the voice <strong>AI</strong> technology used in Taco Bell’s drivethru. They partnered with Taco Bell to implement a conversational AI system that can take orders via voice. The AI uses SoundHound’s <strong>Houndify</strong> platform to recognize and understand customer orders, making the ordering process more efficient and automated.</p><p>I had <strong><em>my first encounter</em></strong> with SoundHound’s AI at a local TacoBell yesterday. The initial results are about what you’d expect. Further training is clearly required, but what surprised me most is that there seems to be zero problems with #ASR (Automatic Speech Recognition). That’s the good news.</p><p>The bad news is the AI’s tempo was set to 11. Way too fast, so it ‘spams’ any milli-silence… with an upsell. For example, the driver in front of me ordered the Chicken Nuggets, and the AI promptly informed her that ‘This location is out of Chicken Nuggets’.</p><p>But instead of giving her time to look at the menu for a substitution, the AI started barking up all kinds of un-helpful suggestions.</p><p>You could say, the AI actually… Hounds.. you.</p><p>I correctly intuited my order would be hard on the AI. I love this <strong>vegetarian</strong> hack on the Crunchy Taco Supremes — ordering Black Beans instead of meat. They’re surprisingly good. The AI fumbled once, but eventually did recover the first order. But then it totally failed trying to add a second REGULAR Taco Supreme (for my passenger).</p><blockquote>As I was considering adding one of the 7 different kinds of quesadilla, the AI just put me on full-blast with suggestions. I rudely told it to <strong><em>“Stop being so damn obnoxious, and give me a few moments!”</em></strong> This made the AI stop completely, and I almost felt bad. For a second. Then I heard the sound of laughter…</blockquote><p>I could see the crew in the Pay Window were all laughing. And a human fast food worker cut-in over the intercom. He very politely apologized for the AI, and completed the order. Properly.</p><p>The crew <strong><em>does</em></strong> like the new AI. They tell me it works well for their ‘Heavy Users’, i.e., folks that really know the menu. But it’s not so great for slow, picky occaisonal users like me, and apparently… it simply can’t handle the clatter of Diesel engines. At all.</p><p>But they expect that with training, it will get (much) better.</p><p>I have just two suggestions for the Bell:</p><p>1) Introduce a conversation behavior model that is not based on simple time-of-response. I can help you with this if you like. But I’m serious. Do it!</p><p>2) Please! Just stop begging us to RoundUp’ for charity. It’s no longer Christmas time, and it’s bad enough when a human asks, but this AI Nagging is especially hideous.</p><p><strong>CONCLUSION:</strong> I own SoundHound stock and my initial experience here made me glad that I do. The stock has done well since I bought it. But it’s now richly valued, and you’ll have to carefully consider buying NASDAQ:HOUN as a very long term play, indeed.</p><p>© 2025 FRED</p><p>Activate to view larger image,</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/800/0*q98ocvKlfaScm0h4" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=28c09791f8ab" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>