Vue lecture

Il y a de nouveaux articles disponibles, cliquez pour rafraîchir la page.
🔲 ⭐

The Personal AI Greenfield

What forms of pAI—personal AI—are Apple, Mozilla, Google, Meta, Microsoft and the rest not doing?

Let’s look at those first two because they’re at the top of the news LIFO buffer.

Apple Intelligence (“coming in beta this fall*“), announced yesterday, will help you with writing and creating images while giving you less lame answers from Siri. (Which they should re-name. Siri is Apple’s Clippy.) It “can draw on larger server-based models, running on Apple silicon, to handle more complex requests for you while protecting your privacy.” The “larger models” will be white-labeled ChatGPT, plus Apple’s own small language models (SLMs).

Mozilla, which got $400+ million a year from Google (for search in the Firefox browser) starting in 2020, announce on June 3 that they will be Building open, private AI with the Mozilla Builders Accelerator. Jive:

This program is designed to empower independent AI and machine learning engineers with the resources and support they need to thrive. It aims to cultivate a more innovative AI ecosystem, and it’s one of Mozilla’s key initiatives to make AI meaningfully impactful — alongside efforts like Mozilla.ai, the Responsible AI Challenge and the Rise25 Awards.

The Mozilla Builders Accelerator’s inaugural theme is local AI, which involves running AI models and applications directly on personal devices like laptops, smartphones, or edge devices rather than depending on cloud-based services…

We chose Local AI as the theme for the Accelerator’s first cohort because it aligns with our core values of privacy, user empowerment, and open source innovation. This method offers several benefits including:

  • Privacy: Data stays on the local device, minimizing exposure to potential breaches and misuse.
  • Agency: Users have greater control over their AI tools and data.
  • Cost-effectiveness: Reduces reliance on expensive cloud infrastructure, lowering costs for developers and users.
  • Reliability: Local processing ensures continuous operation even without internet connectivity.

Looks to me like both of these are Big AI writ small. It’s “local,” not personal. It’s made to serve your needs with what BigAI offers through APIs. It is still essentially AIaaS (AI as a Service), rather than truly personal AI (pAI): personalized more than personal.

That’s also what I see when I read between the lines at Mozilla’s AI job openings. Take platform engineer. This person will (among other things), “assist in managing and orchestrating workloads across multiple cloud providers.” That’s fine. I’m sure true pAIs will do that too. But most of pAI will be more personal than that. It will deal with the mundanities of your everyday life. Not with coughing up answers that can only come from AIaaSes.

The problem with personalizing AI giant offerings is that they are large language models (LLM) trained on everything that can be crawled on the Internet, plus who knows what else. Not on your truly personal stuff. This is why “prompt engineering” worthy of the noun is ” not for anybody:

Prompt engineering is crucial for deploying LLMs but is poorly understood mathematically. We formalize LLM systems as a class of discrete stochastic dynamical systems to explore prompt engineering through the lens of control theory. We investigate the reachable set of output token sequences $R_y(\mathbf x_0)$ for which there exists a control input sequence $\mathbf u$ for each $\mathbf y \in R_y(\mathbf x_0)$ that steers the LLM to output $\mathbf y$ from initial state sequence $\mathbf x_0$. We offer analytic analysis on the limitations on the controllability of self-attention in terms of reachable set, where we prove an upper bound on the reachable set of outputs $R_y(\mathbf x_0)$ as a function of the singular values of the parameter matrices. We present complementary empirical analysis on the controllability of a panel of LLMs, including Falcon-7b, Llama-7b, and Falcon-40b. Our results demonstrate a lower bound on the reachable set of outputs $R_y(\mathbf x_0)$ w.r.t. initial state sequences $\mathbf x_0$ sampled from the Wikitext dataset. We find that the correct next Wikitext token following sequence $\mathbf x_0$ is reachable over 97% of the time with prompts of $k\leq 10$ tokens. We also establish that the top 75 most likely next tokens, as estimated by the LLM itself, are reachable at least 85% of the time with prompts of $k\leq 10$ tokens. Intriguingly, short prompt sequences can dramatically alter the likelihood of specific outputs, even making the least likely tokens become the most likely ones. This control-centric analysis of LLMs demonstrates the significant and poorly understood role of input sequences in steering output probabilities, offering a foundational perspective for enhancing language model system capabilities.

But all that stuff applies mostly when we’re prompting a big LLM system.

What about using AI in our own lives, where the data that matters most are in our calendars, contacts, financial and health records, our travels, our correspondence (email, chat, whatever)? And how about all the location data we might get from our cars, phone apps, and phone companies? These should be much easier for a pAI to gather, examine, and help us do useful things. Caring about much less data also means a pAI will be less likely to give wrong (hallucinated) answers.

Today the mental frame almost everybody uses for AI is the Big kind, ingesting everything they can get their crawlers on, and munching all of it in giant compute farms. Those systems are great for lots of stuff, but they still don’t deal with personal data listed in the last paragraph.

Not yet, anyway.

Look at it this way. For each of us, there are three data pools:

  1. The entire Net, which is what gets crawled by all the giant LLM operators, plus whatever else they can get their claws on.
  2. One’s personal life, some of which is digitized in useful form (contacts, calendar, mail, stuff in folders inside PCs and attached drives).
  3. Personal data that is in the hands of giants, but is rightfully ours. These include our driving record and driving practices (,recorded by our late model cars and snitched to insurance companies and others), our location data (kept and shared by car and phone carriers to the likes of Google and the feds), our TV viewing habits, (gathered by Google, Amazon, Roku, Apple, etc.).

The pAI greenfield is with the last two.

Tell us who is working on what there, preferably with open source, and not sitting on walled garden silicon.

[Later… ] Since readers told me I had small language models (SLMs) wrong in one of the paragraphs above, and I’m not sure I had them right, I rewrote them out of the piece. I invite readers to post comments to further correct and expand on the subject of pAIs and what they can do.

🔲 ☆

Coming soon to a radio near you: Personalized ads

And privacy be damned.

See, there is an iron law for every new technology: What can be done will be done. And a corollary that says, —until it’s clear what shouldn’t be done.  Let’s call those Stage One and Stage Two.

With respect to safety from surveillance in our cars, we’re at Stage One.

For Exhibit A, read what Ray Schultz says in Can Radio Time Be Bought With Real-Time Bidding? iHeartMedia is Working On It:

HeartMedia hopes to offer real-time bidding for its 860+ radio stations in 160 markets, enabling media buyers to buy audio ads the way they now buy digital.

“We’re going to have the capabilities to do real-time bidding and programmatic on the broadcast side,” said Rich Bressler, president and COO of iHeart Media, during the Goldman Sachs Communacopia + Technology Conference, according to Radio Insider.

Bressler did not offer specifics or a timeline. He added: “If you look at broadcasters in general, whether they’re video or audio, I don’t think anyone else is going to have those capabilities out there.”

“The ability, whenever it comes, would include data-infused buying, programmatic trading and attribution,” the report adds.

The Trade Desk lists iHeart Media as one of its programmatic audio partners.

Audio advertising allows users to integrate their brands into their audiences’ “everyday routines in a distraction-free environment, creating a uniquely personalized ad experience around their interests,” the Trade Desk says.

The Trade Desk “specializes in real-time programmatic marketing automation technologies, products, and services, designed to personalize digital content delivery to users.” Translation: “We’re in the surveillance business.”

Never mind that there is negative demand for surveillance by the surveilled. Push-back has been going on for decades.  Here are 154 pieces I’ve written on the topic since 2008.

One might think radio is ill-suited for surveillance because it’s an offline medium. Peopler listen more to actual radios than to computers or phones. Yes, some listening is online; but  not much, relatively speaking. For example, here is the bottom of the current radio ratings for the San Francisco market:

Those numbers are fractions of one percent of total listening in the country’s most streaming-oriented market.

So how are iHeart and The Trade Desk going to personalize radio ads?  Well, here is a meaningful excerpt from iHeart To Offer Real-Time Bidding For Its Broadcast Ad Inventory, which ran earlier this month at Inside Radio:

The biggest challenge at iHeartMedia isn’t attracting new listeners, it’s doing a better job monetizing the sprawling audience it already has. As part of ongoing efforts to sell advertising the way marketers want to transact, it now plans to bring real-time bidding to its 850 broadcast radio stations, top company management said Thursday.

“We’re going to have the capabilities to do real-time bidding and programmatic on the broadcast side,” President and COO Rich Bressler said during an appearance at the Goldman Sachs Communacopia + Technology Conference. “If you look at broadcasters in general, whether they’re video or audio, I don’t think anyone else is going to have those capabilities out there.”

Real-time bidding is a subcategory of programmatic media buying in which ads are bought and sold in real time on a per-impression basis in an instant auction. Pittman and Bressler didn’t offer specifics on how this would be accomplished other than to say the company is currently building out the technology as part of a multi-year effort to allow advertisers to buy iHeart inventory the way they buy digital media advertising. That involves data-infused buying and programmatic trading, along with ad targeting and campaign attribution.

Radio’s largest group has also moved away from selling based on rating points to transacting on audience impressions, and migrated from traditional demographics to audiences or cohorts. It now offers advertisers 800 different prepopulated audience segments, ranging from auto intenders to moms that had a baby in the last six months…

Advertisers buy iHeart’s ad inventory “in pieces,” Pittman explained, leaving “holes in between” that go unsold. “Digital-like buying for broadcast radio is the key to filling in those holes,” he added…

…there has been no degradation in the reach of broadcast radio. The degradation has been in a lot of other media, but not radio. And the reason is because what we do is fundamentally more important than it’s ever been: we keep people company.”

Buried in that rah-rah is a plan to spy on people in their cars. Because surveillance systems are built into every new car sold. In Privacy Nightmare on Wheels’: Every Car Brand Reviewed By Mozilla — Including Ford, Volkswagen and Toyota — Flunks Privacy Test, Mozilla pulls together a mountain of findings about just how much modern cars spy on their drivers and passengers, and then pass personal information on to many other parties. Here is one relevant screen grab:

spying

As for consent? When you’re using a browser or an app, you’re on the global Internet, where the GDPR, the CCPA, and other privacy laws apply, meaning that websites and apps have to make a show of requiring consent to what you don’t want. But cars have no UI for that. All their computing is behind the dashboard where you can’t see it and can’t control it. So the car makers can go nuts gathering fuck-all, while you’re almost completely in the dark about having your clueless ass sorted into one or more of Bob Pittman’s 800 target categories. Or worse, typified personally as a category of one.

Of course, the car makers won’t cop to any of this. On the contrary, they’ll pretend they are clean as can be. Here is how Mozilla describes the situation:

Many car brands engage in “privacy washing.” Privacy washing is the act of pretending to protect consumers’ privacy while not actually doing so — and many brands are guilty of this. For example, several have signed on to the automotive Consumer Privacy Protection Principles. But these principles are nonbinding and created by the automakers themselves. Further, signatories don’t even follow their own principles, like Data Minimization (i.e. collecting only the data that is needed).

Meaningful consent is nonexistent. Often, “consent” to collect personal data is presumed by simply being a passenger in the car. For example, Subaru states that by being a passenger, you are considered a user — and by being a user, you have consented to their privacy policy. Several car brands also note that it is a driver’s responsibility to tell passengers about the vehicle’s privacy policies.

Autos’ privacy policies and processes are especially bad. Legible privacy policies are uncommon, but they’re exceptionally rare in the automotive industry. Brands like Audi and Tesla feature policies that are confusing, lengthy, and vague. Some brands have more than five different privacy policy documents, an unreasonable number for consumers to engage with; Toyota has 12. Meanwhile, it’s difficult to find a contact with whom to discuss privacy concerns. Indeed, 12 companies representing 20 car brands didn’t even respond to emails from Mozilla researchers.

And, “Nineteen (76%) of the car companies we looked at say they can sell your personal data.”

To iHeart? Why not? They’re in the market.

And, of course, you are not.

Hell, you have access to none of that data. There’s what the dashboard tells you, and that’s it.

As for advice? For now, all I have is this: buy an old car.

 

 

❌