Part 1: For decades, tobacco companies claimed that smoking wasn’t harmful. Did they have proof? Of course not. Did the oil industry have any proof that it was safe to keep using lead as an additive in gasoline? Nope. The claimed health benefits of drinking red wine didn’t hold up to impartial scrutiny.
Do deniers of climate change have a valid point in their refusal to accept Al Gore’s inconvenient truth? No, they do not, yet they persist as if a few, isolated papers written by partisan cranks have enough weight to counter the conclusions of almost every climate expert. Using AI to amplify lies as if they are validated facts isn’t just a possibility, it’s already happening.
I don’t keep up with technology as much as I did when working in the business, but it appears the Hadoop big data project is no longer the next big thing. Now it’s the Large Language Model (LLM). The results generated by Artificial Intelligence searches follow the old truism, “garbage in, garbage out.”
Robots.txt: “A robots.txt file tells search engine crawlers which URLs the crawler can access on your site.”
https://developers.google.com/search/docs/crawling-indexing/robots/intro
XML Sitemap: “A sitemap is a file where you provide information about the pages, videos, and other files on your site, and the relationships between them.”
https://developers.google.com/search/docs/crawling-indexing/sitemaps/overview
LLM: “… a statistical language model, trained on a massive amount of data, that can be used to generate and translate text and other content…”
https://cloud.google.com/ai/llms
LLMS.txt: “A proposal to standardise [sp] on using an /llms.txt file to provide information to help LLMs use a website at inference time.”
https://llmstxt.org/
Although robots.txt isn’t a mandatory file, it’s safe to say almost every website has one. Creating an XML Sitemap requires a bit more work. I didn’t have one until getting serious about fixing the crippling technical problem (resolved in December, 2021) that kept this place from appearing on Google and prevented WordPress from updating itself.
Do we need another method of optimizing web searches? LLMS.txt has the potential of being misused by intentionally manipulating the Large Language Model for AI in ways that could validate false or misleading data.
Bitcoin started the rush for Nvidia graphic processors that denied PC gamers their highest framerate FPS fun. Then AI further accelerated the demand for GPU’s. So both technologies are relying on the same class of hardware.
For all of the emphasis on blockchain conferring incontrovertible provenance of a Bitcoin’s digital token ownership, where is the assurance of provenance for AI’s “knowledge”? I wanted to put AI to a very small, innocuous and perhaps irrelevant test.
Intermission:
Part 2: A particular class of technology that I have been more actively keeping up with is video projectors.
TI’s DLP projector technology is ideal for movie theaters, where light loss must be minimal. Professional DLP projectors have three mirror array chips. 3LCD projectors are limited mostly to home and office use. They have, as the name implies, three LCD panels, one for each primary additive color, being red, green, and blue.
Home DLP projectors have a single chip. That means the colors must be projected sequentially. Although single-chip DLP has no convergence error like 3LCD, and its contrast ratio and black level are superior to 3LCD, for a long time there was no comparison between DLP’s relatively muted color output and 3LCD’s superior color quality.
A significant issue for me with single-chip DLP is the “rainbow effect.” Perhaps it’s due to my cataract replacement lenses, but I am extremely susceptible to the rainbow effect. As documented thirteen years ago, I was likewise affected by “phosphor lag” on plasma TV’s.
Triple laser, sometimes combined with LED, is a new way of generating color in single-chip DLP projectors. It eliminates the old, mechanical spinning color wheel. Triple laser is finding its place in short-throw projectors, a type of product that holds no interest for me.
The result with triple lasers compared to the old color wheel is outstanding color reproduction. But is the rainbow effect eliminated? Techmoan takes a while to bring it up in his review of the Nebula Cosmos 4K SE projector (not a short-throw model), but he gets there.
Two years ago I bought a similar product, the Epson Mini EF12, a 3LCD projector. Techmoan says the Nebula’s black level can’t compare with an OLED TV, but nothing can except for a still-working plasma TV. The native contrast for an inexpensive 3LCD projector like the EF12 may be only 400:1, whereas a comparable triple laser DLP could be 3500:1. (The ratings for JVC D-ILA high-end projectors start at 40,000:1.)*
The Nebula’s software settings are far more advanced than the limited controls on the Epson. As impressive as the Nebula projector is for the money, all I needed to hear Techmoan saying was, “The rainbow effect is sometimes noticeable on the projector.” No thanks.
The EF12 is my projector for casual viewing in the Pratt Cave. It isn’t suitable for dark scenes, because they appear very washed out. Cartoons and brightly lit news broadcasts look quite good, as do many YouTube videos, like the ones posted by Techmoan.
So what does all of this have to do with AI? Thanks to Techmoan I know DLP exhibits the rainbow effect, even with a triple laser. Now I want to know why that’s true. I used Chrome to search Google and specified AI Mode:
“why do triple laser dlp projectors have rainbow effect”
The explanation for the rainbow effect, that “the lasers themselves rapidly switch between the primary colors,” seems to come from some guy on Reddit. Is it correct? My little innocuous search reveals that a plausible answer has been obtained from an unknown source. All I know for certain is Techmoan can see the rainbow and he doesn’t mind it, but it would distract me to the point of getting a headache.
Part 3: AI definitely has a future in medicine. Although the risks for a patient can be great in the event of a mistake, there’s so much potential in AI to improve diagnosis and the quality of medical imaging. AI being able to sift through millions of confirmed cases — anonymously, presumably — could greatly improve the quality of care. For example, the symptoms of conditions such as Parkinson’s Disease and colon cancer wouldn’t be ignored or misdiagnosed by a relatively inexperienced physician who might otherwise be inclined to tell a patient, “you’re too young to have [disease name].”
AI used for partisan political purposes scares me, and it isn’t reassuring to know that it can be used effectively by both sides. The genie is out of the bottle, but be careful what you ask for. I say, “everybody slow down and take a breath.”
* To my surprise, Techmoan’s DLP projector has a rated contrast ratio of only 400:1.
