Category: AI

Exploring Airport Security: How Baggage Scanners Work | September 02 2025, 20:29

The day after tomorrow, I am flying to Amsterdam (and then to Turkey), and I remembered that I had an unanswered question to myself about how baggage scanners work at the airport. Of course, I knew that it was essentially computer tomography, X-rays and all that, but I wanted more details. And below is the response as to why they ask you to take out water, and why sometimes they do not.

It turns out that modern scanners can not only see the shape of objects but also determine what material they are made of. How does a regular scanner work? Dense materials (such as metal) absorb a lot of radiation and appear bright or opaque in images. Less dense materials absorb little radiation and appear dark. Hence laptops, for example, had to be taken out — not because the scanner couldn’t recognize them, but because their dense components (battery, boards) could be used to hide other prohibited items behind them. So, it has long been not just scanners, but computer tomography — in essence, the bag or suitcase is scanned from all sides, then a 3D image is created. It seems like everyone knows this.

But I mentioned that they understand the materials items are made from. How?

It turns out that the scanner uses dual-energy X-ray technology. It scans the object with two beams of rays of different energy levels (high and low). Since materials absorb radiation differently depending on the energy of the ray and their atomic composition, the system analyzes this difference. Based on the absorption ratio of the two beams, the effective atomic number Z — a key characteristic, a kind of “elemental fingerprint” of the substance, is calculated.

The problem is that this “fingerprint” of water (~7.4) and many explosives are almost identical. This is precisely why water was banned. Relying only on this parameter would mean receiving a huge number of false alarms.

Here is where computer tomography (CT) comes into play. The scanner creates an accurate three-dimensional (3D) model of the contents of the bag. From the 3D model, the system obtains the exact volume (V) of each object. Based on data on the absorption of X-rays, its mass (m) is calculated. Then it’s simple: ρ=m/V.

That is, the system does not make a decision based on one parameter. It plots each detected substance on a two-dimensional graph with axes “Z — density.” On this graph, water and explosives, having almost the same atomic number, occupy completely different positions due to different densities.

And that’s precisely why water can sometimes be carried through. Smart machines simply do not mark it as something significant, but still identify it as water. Then procedures follow. If the airport has updated the machines, but not the procedures, they will ask to dispose of the water. But also, not all machines are updated everywhere, and at the same airport, it depends on which line is open at the moment.

The cost of such a scanner is $300-400 thousand.

The scanners for people work differently. They use millimeter waves. They pass through clothing and reflect back from the skin. Water absorbs them significantly, so they penetrate only a couple of millimeters. The system registers the reflected signal and constructs a three-dimensional map of the body surface and objects under the clothing. But it does not show this — instead, it displays a simplified contour of a person and shows on it what ML found unusual. Therefore, by the way, many try to carry various items inside themselves, knowing that such a scanner absolutely cannot see it.

Scam Alert: The Bond Ring Energy Hoax Following My Oura Ring Purchase | August 20 2025, 20:01

I had just bought the Oura Ring 4 when Facebook started running scam ads about the first ring that saps your energy for its own survival. My precious!..

Misguided Lessons with Grok: A Bilingual Blunder | August 19 2025, 23:43

Today Grok blew my mind. I say, teach me French. He says, ok, how do you say “book”? I say “le livre”. He says “wrong! la livra”. 😳The car drives itself anyway, decided to record the dialogue. He’s not convinced. At all, insists on his point. La livra and that’s it. I’m afraid Grok will teach the bad stuff in his Language Tutor mode.

I remembered a story from “Memoirs of Pushkin” by M. E. Yuzefovich, dating to 1829:

he had several books with him, including Shakespeare. One day in our tent, he translated some scenes to me and my brother. I had once studied English, but having not fully learned it, I subsequently forgot it. However, I still recognized its sounds. In Pushkin’s reading, the English pronunciation was so distorted that I suspected his knowledge and decided to test it. The next day, I invited his relative, Zakhar Chernyshev, who knew English as his native language, warned him what was going on, and called over Pushkin with Shakespeare. He willingly started translating for us. Chernyshev burst into laughter at the first words read by Pushkin: “First tell me, in which language are you reading?” Pushkin laughed in turn, explaining that he had taught himself English, and therefore he reads English letters like Latin ones. But the fact is that Chernyshev found the translation completely correct and the language understanding impeccable.”

Anna Derevenitskaya

Exploring AI Search Agent: Revolutionizing Automated Browsing and Task Completion | August 19 2025, 01:21

In addition to the main product for search testing, I am developing an AI Search Agent in my leisure time. You only need to provide it with two pieces of information: a website to visit and a goal (described in a short paragraph). In other words, this thing is smart enough to function without any setup – just the site and the goal, and then it’s on its own.

How it works: This virtual agent generates search queries on its own, refines them based on the results obtained (for example, simplifies them), and analyzes how well they match the intended purpose. If suitable results are found, the agent can add items to the cart and place an order — if this is configured in the settings.

I’ve already written about this recently – today is just a slightly nicer demo. It will be even nicer as it is still being pulled from the middle of development, but you can already see how the page is analyzed, and there are initial results that can be used.

The agent can be used for several purposes. Firstly, it’s an excellent way to create ground truth—a set of queries with perfect results. These data can then be used for search testing without involving often slow and expensive large language models (LLM). Secondly, it helps to test the search functions before deploying them to users. Thirdly, the agent generates realistic usage data needed for training recommendation models that require authentic interactions.

The colorful rectangles in the video are the language of interaction of the agent with AI (or LLM). To understand where to click, the system annotates the page and sends a structured description of the page to AI—often along with a screenshot—so it can analyze everything and make a decision about the next action.

Exploring TestMySearch.com’s Virtual Shopper System | August 15 2025, 04:27

As part of the TestMySearch.com project, I am creating a “virtual shopper” system that simulates the behavior of a real user in an online store: it starts with an abstract goal (for example, “something bright and sexy for the gym”), turns it into a specific search query, performs the search on the site, and depending on the results, may either continue browsing or, with a certain probability, reformulate the query if the findings do not match the original goal; the system then evaluates the pages for their alignment with the initial idea, opens product cards, randomly changes parameters such as color or size, makes decisions about adding to the cart and placing an order, and may also leave the site, which allows generating many sessions similar to real ones overnight for testing search, filters, and recommendations even before live users arrive.

The system is fully automatic. That is, the browser in the video opens by itself, the search field appears by itself (i.e., independent of the site), the system itself concocts the text based on that very initial goal, then the facets and search results are displayed, which may also be in a form unpredictable to the system — but it still understands what is what, and makes decisions about whether to rephrase the query, select a facet or click on a search result. There is a certain probability that the virtual user will leave the site. If the query is reformulated, for example, this virtual user does not repeat queries that have already led to empty or irrelevant results, so within the session there is “memory”.

Unlocking Smartwatches with Unique Heart Rhythms: A Missed Opportunity? | August 06 2025, 16:43

Why has no one made it so that smartwatches only unlock on the wrist of their owner, reading their unique heartbeat or other biometric data? This is in addition to having the owner’s phone nearby.

Officially, you can’t disable this in the settings of an Apple Watch — Apple intentionally made it such that when you put on the watch for the first time each day, it always requires a code, even if the iPhone is nearby. This is due to security policy: the watch may end up on someone else’s wrist, and the phone may just be nearby.

Moreover, every person has unique heart rhythm patterns, which include, for example, slight variations in the intervals between heartbeats, characteristics of the heart signal shape, and how the heart responds to different stresses. These microscopic differences create a unique picture” of heart rhythm that is difficult to fake or replicate. Watches have quite a lot of time, after being worn and before they are needed unlocked, to collect, process, and decide whether to unlock or not.

Navigating Code Generation with AI: Essential Skills for Programmers | August 04 2025, 14:28

I am currently using Gemini extensively for code generation, and I see a skill that programmers need to have to be successful in this field. It’s the ability to quickly read and understand someone else’s code, as well as explain why AI generation needs to be redone and how. For the former, you simply need to know the language very well and read “from the sheet,” because there will be little time to ponder. For the latter, you need to know patterns well and understand where they apply and where they do not. AI will still mess up using patterns inappropriately for a long time.

Moreover, a person will still need to understand “as a whole” 90% of the code generated by AI, and also manage to find time to comprehend each generated line of code. If you relax and miss it, the system may produce even working, but very poorly maintainable code. For instance, there is an unwritten rule that individual files should not contain so much code, and if it grows, you need to refactor, breaking one large into two or three. Sometimes this requires rewriting logic, but this rewriting is always aimed at one task – to simplify maintenance. And AI, while rewriting, also “improves” the code at the same time. And this is quite difficult to prohibit.

In addition, the very concept of LLM implies the limitation of the contextual window. Which gets filled with code very quickly. To create an illusion for the user that everything is working even with a large volume of code, LLMs are able to do preliminary processing, extracting only relevant pieces for processing and setting aside irrelevant ones, so that the relevant ones fit into the actual contextual window. But this process is very unreliable, and once it works, and the second time it turns out that something important was set aside, and as a result, the system did not see the whole picture and generated code, which includes a function very similar to the function set aside, and now we have two almost identical ones.

Besides, currently logic is distributed between the DB and the code. That is, data often controls the code. And data in LLMs simply often do not fit. There is too much of it. In the end, without programmers, current LLM architectures cannot cope. But the requirements for programmers’ qualifications will only increase with LLMs, not decrease. So yes, juniors should be worried, but leads not so much 🙂

Seattle Airport Chaos: IT Glitches and Alaska Airlines Grounded | July 21 2025, 07:07

Seattle Airport is at a standstill – some nonsense with IT systems, Alaska Airlines planes are not taking off (grounded).

UPDATE: remember the door that fell off Boeing mid-flight? It was Alaska Airlines and Boeing 737 Max, which I am currently sitting in.

Why Aren’t Smart Systems Widely Used in Commercial Vehicles? | July 18 2025, 20:33

I wonder why smart systems, cameras, driver assistance systems in driving are not used on commercial transport such as trucks and buses? It’s one thing to integrate such statuses into a $35K car, and another into a truck or bus, whose prices start at least at $100-150K, and in some cases more. Buses are often purchased by organizations for whom an extra $5-10K on a price of $100-150 may not make much of a “difference” in deciding what to buy. Although of course understanding that there, with a tender for the minimum price, every thousand could be decisive. On the other hand, insurance might be lower, and it can be nicely sold to people (passengers). Also, it seems that truck drivers falling asleep are simply more dangerous and costly than personal car drivers falling asleep.

Why Don’t We Have Self-Sustaining Solar-Powered Drones Yet? | July 16 2025, 01:33

I wonder why we still don’t see autonomous drones that could lead an “eternal” life: landing on roofs, deploying solar panels, charging from the sun, and taking off once a day for whatever their mission might be? When you consider the energy aspects, it seems like a feasible scheme. For instance, a heavy drone weighing about 8 kg could carry foldable solar panels with an area of 1.5 m² and a battery with a capacity of 2 kWh. In one sunny day, such panels could collect about 1.2 kWh of energy — enough for it to fly for 20 minutes at a speed of 40–50 km/h, take photos, and transmit them via the mobile network. And there would still be a reserve of energy for several cloudy days.

Even a light drone weighing 2 kg with small panels (0.5 m²) could rise into the air for 10–15 minutes every day if it managed to find good weather and a sunny roof. The power required for hovering for such devices is about 150–200 W, and solar panels with 20% efficiency at mid-latitudes can produce up to 350–400 Wh per day. The balance comfortably adds up, especially if not chasing speed and if there’s no rush on the roof.

Such a “solar nomad” could live for weeks and months, flying from roof to roof and charging in anticipation of missions. At first glance, the technology of batteries and panels already allows this to be done. Or am I missing something?