AI Delivers The Best of Both Worlds: Deception and Inaccuracy

May 16, 2024

dinosaur30a_thumbThis essay is the work of a dinobaby. Unlike some folks, no smart software improved my native ineptness.

Wizards from one of Jeffrey Epstein’s probes made headlines about AI deception. Well, if there is one institution familiar with deception, I would submit that the Massachusetts Institute of Technology might be considered for the ranking, maybe in the top five.

The write up is “AI Deception: A Survey of Examples, Risks, and Potential Solutions.” If you want summaries of the write up, you will find them in The Guardian (we beg for dollars British newspaper) and Science Alert. Before I offer my personal observations, I will summarize the “findings” briefly. Smart software can output responses designed to deceive users and other machine processes.


Two researchers at a big name university make an impassioned appeal for a grant. These young, earnest, and passionate wizards know their team can develop a lie detector for an artificial intelligence large language model. The two wizards have confidence in their ability, of course. Thanks, MSFT Copilot. Good enough, like some enterprise software’s security architecture.

If you follow the “next big thing” hoo hah, you know that the garden variety of smart software incorporates technology from outfits like Google. I have described Google as a “slippery fish” because it generates explanations which often don’t make sense to me. Using the large language model generative text systems can yield some surprises. These range from images which seem out of step with historical fact to legal citations that land a lazy lawyer (yes! alliteration) in a load of lard.

The MIT researcher has verified that smart software may emulate the outstanding ethical qualities of an engineer or computer scientist. Logic is everything. Ethics are not anything.

The write up says:

Deception has emerged in a wide variety of AI systems trained to complete a specific task. Deception is especially likely to emerge when an AI system is trained to win games that have a social element …

The domain of the investigation was games. I want to step back and ask, “If LLMs are not understood by their developers, how do we know if deception is hard wired into the systems or that the systems learn deception from their developers with a dusting of examples from the training data?”

The answer to the question is, “At this time, no one knows how these large-scale systems work. Even the “small” LLMs can prove baffling. We input our own data into Mistral and managed to obtain gibberish. Another go produced a system crash that required a hard reboot of the Mac we were using for the test.

The reality appears to be that probability-based systems do not follow the same rules as a human. With more and more humans struggling with old-school skills like readin’, writin’ and ‘rithmatic  — most people won’t notice. For the top 10 percenters, the mistakes are amusing… sometimes.

The write up concludes:

Training models to be more truthful could also create risk. One way a model could become more truthful is by developing more accurate internal representations of the world. This also makes the model a more effective agent, by increasing its ability to successfully implement plans. For example, creating a more truthful model could actually increase its ability to engage in strategic deception by giving it more accurate insights into its opponents’ beliefs and desires. Granted, a maximally truthful system would not deceive, but optimizing for truthfulness could nonetheless increase the capacity for strategic deception. For this reason, it would be valuable to develop techniques for making models more honest (in the sense of causing their outputs to match their internal representations), separately from just making them more truthful. Here, as we discussed earlier, more research is needed in developing reliable techniques for understanding the internal representations of models. In addition, it would be useful to develop tools to control the model’s internal representations, and to control the model’s ability to produce outputs that deviate from its internal representations. As discussed in Zou et al., representation control is one promising strategy. They develop a lie detector and can control whether or not an AI lies. If representation control methods become highly reliable, then this would present a way of robustly combating AI deception.

My hunch is that MIT will be in the hunt for US government grants to develop a lie detector for AI models. It is also possible that Harvard’s medical school will begin work to determine where ethical behavior resides in the human brain so that can be replicated in one of the megawatt munching data centers some big tech outfits want to deploy.

Four observations:

  1. AI can generate what appears to be “accurate” information, but that information may be weaponized by a little-understood mechanism
  2. “Soft” human information like ethical behavior may be difficult to implement in the short term, if ever
  3. A lie detector for AI will require AI; therefore, how will an opaque and not understood system be designated okay to use? It cannot at this time
  4. Duplicity may be inherent in the educational institutions. Therefore, those affiliated with the institution may be duplicitous and produce duplicitous content. This assertion raises the question, “Whom can one trust in the AI development chain?

Net net: AI is hot because is a candidate for 2024’s next big thing. The “big thing” may be the economic consequences of its being a fairly small and premature thing. Incubator time?

Stephen E Arnold, May 16, 2024

Cow Control or Elsie We Are Watching

April 1, 2024

green-dino_thumb_thumb_thumbThis essay is the work of a dumb dinobaby. No smart software required.

Australia already uses remotely controlled drones to herd sheep. Drones are considered more ethnical tax traditional herding methods because they’re less stressful for sheep.

Now the island continent is using advanced tracking technology to monitor buffalos and cows. Euro News investigates how technology is used in the cattle industry: “Scientists Are Attempting To Track 1000 Cattle And Buffalo From Using GPS, AI, And Satellites.”

An estimated 22000 buffalo freely roam in Arnhem Land, Australia. The emphasis is on estimate, because the exact number is unknown. These buffalo are harming Arnhem Land’s environment. One feral buffalo weighing 1200 kilograms and 188 cm not only damages the environment by eating a lot of plant life but also destroys cultural rock art, ceremonial sites, and waterways. Feral buffalos and cattle are major threats to Northern Australia’s economy and ecology.

Scientists, cattlemen, and indigenous rangers have teamed up to work on a program that will monitor feral bovines from space. The program is called SpaceCows and will last four years. It is a large-scale remote herd management system powered by AI and satellite. It’s also supported by the Australian government’s Smart Farming Partnership.

The rangers and stockmen trap feral bovines, implant solar-powered GPS tags, and release them. The tags transmit the data to a space satellite located 650 km away for two years or until it falls off. SpaceCows relies on Microsoft Azure’s cloud platform. The satellites and AI create a digital map of the Australian outback that tells them where feral cows live:

“Once the rangers know where the animals tend to live, they can concentrate on conservation efforts – by fencing off important sites or even culling them. ‘There’s very little surveillance that happens in these areas. So, now we’re starting to build those data sets and that understanding of the baseline disease status of animals,’ said Andrew Hoskins, a senior research scientist at CSIRO.

If successful, it could be one of the largest remote herd management systems in the world.”

Hopefully SpaceCows will protect the natural and cultural environment.

Whitney Grace, April 1, 2024

Google: Technology Which Some Day Will Be Error Corrected But In the Meantime?

May 20, 2021

I read “Can Google Really Build a Practical Quantum Computer by 2029?” which is based on the Google announcements at its developers’ conference. The article reports:

… the most interesting news we heard at I/O has to be the announcement that Google intends to build a new quantum AI center in Santa Barbara where the company says it will produce a “useful, error-corrected quantum computer” by 2029.

Didn’t Google announce “quantum supremacy” some time ago? This is an assertion which China appears to dispute. See “Google and China Duke It Out over Quantum Supremacy.” Let’s assume Google is the supremo of quantumness. It stands to reason that the technology would be tamed by Googlers. With smart software and quantum supremacy, what’s with the 96-month timelines for “a useful” quantum computer?

Then I came across a news item from Fox10 in Arizona. The TV station’s story “Driverless Waymo Taxi Gets Stuck in Chandler Traffic, Runs from Support Crew” suggests that a Google infused smart auto got confused and then tried to get away from the Googlers who were providing “support.” That’s a very non-Google word “support.” The write up asserts:

A driverless Waymo taxi was caught on camera going rogue on a Chandler intersection near a construction site last week. The company told the passenger at the time that the vehicle was confused with cones blocking a lane.

The Google support team lifted off. Upon arrival, the smart taxi with a humanoid in the vehicle “decided to make for a quick getaway.”

According to Business Insider, “Waymo issued a statement that it has assessed the event and used it to improve future rides.” If you want to watch smart software and the incident, you can navigate to this YouTube link (YouTube videos can be disappeared, so no guarantees that the video will remain on the Googley service.)

Amusing. Independent, easily confused smart Google vehicles and error-corrected quantum computers. Soon. Perhaps both the Waymo capabilities and the quantum supremacy are expensive high school science club experiments which may not work in way that the hapless rider in the errant Waymo taxi would describe as “error corrected”?

Stephen E Arnold, May 20, 2021

Smart Software Names Cookies

December 11, 2018

Tired of McVities’ digestives, coconut macaroons, and chocolate chip cookies. Tireless researchers have trained smart software to name cookies. Next to solving death, this is definitely a significant problem.

The facts appear in “AI System Tries to Rename Classic Cookies and Fails Miserably.” You can read the original, possible check the cv of the expert who crafted this study, and inform your local patisserie that you want new names for the confections. (I assume the patisserie has not been trashed by gilets jaunes.)

Here’s an alphabetical list of the “new” names from the write up. Sorry, I don’t have the real world cookie name to which each neologism is matched. Complain to TechRadar, not me.

The names:

  • Apricot Dream Moles
  • Canical Bear-Widded Nuts
  • Fluffin coffee drops
  • Granma’s spritches
  • Hersel pump sprinters
  • Lord’s honey fight
  • Low fuzzy feats
  • Merry hunga poppers
  • Quitterbread bars
  • Sparry brunchies #2
  • Spice biggers
  • Walps

And my personal favorite:

Hand buttersacks.

I quite like the system. One can use it to name secret projects. I can envision attending a meeting and suggesting, “Our new project will be code named Quitterbread bars.”

Stephen E Arnold, December 11, 2018

Knowledge Supposedly the Best Investment

December 13, 2017

Read, read, read, read!  You are told it is good for you, but, much like eating vegetables, no one wants to do it.  School children loath their primers, adults say they do not have the time, and senior citizens explain it puts them to sleep.  Reading, however, is the single best investment an individual can make.  This is not new, but the Observer treats reading like some epiphany in the article, “If You’re Not Spending Five Hours Per Week Learning, You’re Being Irresponsible.”

The article opens with snippets about famous smart people and how they take the time to read at least an hour a day.  The stories are followed by these wise words:

The answer is simple: Learning is the single best investment of our time that we can make. Or as Benjamin Franklin said, ‘An investment in knowledge pays the best interest.’  This insight is fundamental to succeeding in our knowledge economy, yet few people realize it. Luckily, once you do understand the value of knowledge, it’s simple to get more of it. Just dedicate yourself to constant learning.

The standard excuse follows that in today’s modern world we are too busy making money in order to survive to learn new things, then we are slugged with the dire downer that demonetization is making previously expensive technology cheaper or even free.  Examples are provided such as video conferencing, video game consoles, cameras, encyclopedias, and anything digital.  All of these are found on a smartphone.

Technology that was once gold is now cheap, making knowledge more valuable.  Then we are told that technology will make certain jobs obsolete and the only way to survive in the future will be to gain more knowledge and apply, because this can never be taken from you. The bottom line is to read, learn, apply knowledge, and then make that a daily ritual.  The message is not anything new, but does learning via filtered and censored online search results count?

Whitney Grace, December 13, 2017

eDiscovery to Get a Fillip with DISCO

June 23, 2017

A legal technology company has unveiled the next generation AI platform that will reduce time, efforts and money spent by law firms and large corporations on mundane legal discovery work. Named DISCO, the program was in beta phase for two years.

PR distribution platform BusinessWire in a release titled DISCO Launches Artificial Intelligence Platform for Legal Technology quotes:

While there will be many applications for DISCO AI, initially the focus is to dramatically reduce the time, burden, and cost of identifying evidence in legal document review — a process known as eDiscovery.

Many companies have attempted to automate the process of eDiscovery, the success rates, however, have been far from encouraging. Apart from disrupting the legal industry, automated processes like the ones offered by DISCO will render many people in the industry jobless. AI creators, however, say that their intention is to speed up the process and reduce costs to organizations. But again, as technology advances, job losses are inevitable.

Vishal Ingole,  June 23, 2017

Make Your Amazon Echo an ASMR Device

June 7, 2017

For people who love simple and soothing sounds, the Internet is a boon for their stimulation.  White noise or ambient noise is a technique many people use to relax or fall asleep.  Ambient devices used to be sold through catalogs, especially Sky Mall, but now any sort of sound can be accessed through YouTube or apps for free.  Smart speakers are the next evolution for ambient noise.  CNET has a cool article that explains, “How To Turn Your Amazon Echo Into A Noise Machine.”

The article lists several skills that can be downloaded onto the Echo and the Echo Dot.  The first two suggestions are music skills: Amazon Prime Music and Spotify.  Using these skills, the user can request that Alexia finds any variety of nature sounds and then play them on a loop.  It takes some trial and error to find the perfect sounds to fit your tastes, but once found they can be added to a playlist.  An easier way, but might offer less variety is:

One of the best ways to find ambient noise or nature sounds for Alexa is through skills. Developer Nick Schwab created a family of skills under Ambient Noise. There are currently 12 skills or sounds to choose from:

  • Airplane

  • Babbling Brook

  • Birds

  • City

  • Crickets

  • Fan

  • Fireplace

  • Frogs

  • Ocean waves

  • Rainforest

  • Thunderstorms

  • Train

Normally, you could just say, “Alexa, open Ambient Noise,” to enable the skill, but there are too many similar skills for Alexa to list and let you choose using your voice. Instead, go to or open the iOS or Android app and open the Skills menu. Search for Ambient Noise and click Enable.

This is not a bad start for ambient noises, but the vocal command adds its own set of problems.  Amazon should consider upgrading their machine learning algorithms to a Bitext-based solution.  If you want something with a WHOLE lot more variety to check out YouTube and search for ambient noise or ASMR.

Whitney Grace, June 7, 2017

Snapchat Introduces Search Feature

May 29, 2017

Photo-sharing app Snapchat is late to the search game, but it has now arrived. The Daily Mail reports, “Snapchat Introduces a ‘Universal Search’ Feature: Tool Lets You Create Groups and Find New People to Follow.” Writer Abigail Beall explains:

Snapchat’s universal search bar hopes to address an issue some users had with the photograph-sharing app – the difficulty in finding new people to follow and gaining a large following. Previously, the only way people could gain a following was by sharing their username, or Snapcode, outside of the app. The new search bar, that will always be present at the top of the app, will allow people to find users easily through searching, discovering and groups. …


The new feature also lets users create groups, to combine snaps. Previously, boxes for finding specific conversations, accounts to follow and Stories or Discover channels were all in different places.

The tool was implemented for some Android users in mid-January, with availability to all Android and iOS users to follow “soon.” Beall notes the development was predicted by some last August after Snapchat acquired Vurb, a mobile search startup founded in 2011 and based in San Francisco.

Snap Inc., Snapchat’s parent company, bills itself as a camera company that is reinventing the camera. The company has acquired nine other enterprises since its founding in 2011. Snap is now selling (through their special vending machines!)  Spectacles, sunglasses with a camera on each temple that, of course, link right in with Snapchat.

Cynthia Murrell, May 29, 2017

UK Big Brother Invades More Privacy

April 18, 2017

The United Kingdom has been compared to George Orwell’s 1984 dystopia before, especially in the last two decades with their increasing amount of surveillance technology.  Once more UK citizens face privacy invasion reports the Guardian in “UK Public Faces Mass Invasion Of Privacy As Big Data And Surveillance Merge.”  The UK’s Surveillance Camera Commissioner Tony Porter expressed his worry that government regulators were unable to keep up with technological advances.

Big data combined with video surveillance, facial recognition technology, and the profuse use of more cameras is making it harder to protect individuals’ privacy.  People are being recorded 24/7 and often without their knowledge.  Another worry is that police are not being vigilant with private information.  One example is that license plate information has not been deleted after the two-year limit.

Porter wants changes to be made in policies and wants people to be aware of the dangers:

Porter’s new strategy, published on Tuesday, points out that an overwhelming majority of people currently support the use of CCTV in public places. But he questions whether this support can continue because of the way surveillance is changing.


‘I’m worried about overt surveillance becoming much more invasive because it is linked to everything else,’ Porter said. ‘You might have a video photograph of somebody shopping in Tesco. Now it is possible to link that person to their pre-movements, their mobile phone records, any sensor detectors within their house or locality. As smart cities move forward, these are challenges are so much greater for people like myself. And members of the public need to decide whether they are still happy with this.’

Porter admitted that advanced surveillance technology had allowed law enforcement to arrest terrorists and track down missing people, but it still can lead to worse privacy invasions.  Porter hopes is new three-year strategy will inform authorities about how technology will impact privacy.

The good thing about surveillance technology is how it can track down bad guys, but it can be harmful to innocent citizens.  The BBC should run some PSAs about video surveillance and privacy to keep their citizens informed.  I suggest they do not make them as scary as this one about electricity.

Whitney Grace, April 18, 2017

Cortana Becomes an MD

April 17, 2017

Smartphone assistants like Microsoft’s Cortana are only good for verbal Internet searches.  They can be made smarter with an infusion of machine learning and big data.  According to Neowin, Microsoft is adding NLP and AI to Cortana and sending it to medical school, “The UK’s Health Services Now Relies On Cortana Intelligence Suite To Read Medical Research.”

Microsoft takes a lot of flak for their technology, but they do offer comprehensive solutions that do amazing things…when they work.  The UK Health Services will love and hate their new Cortana Intelligence Suite.  It will be utilized to read and catalog medical research to alert medical professionals to new trends in medicine:

Researching and reading can consume medical professionals’ times, stealing a valuable resource from patients.

That’s why the UK’s National Institute for Health and Care Excellence (NICE) is now relying on Microsoft’s Cortana Intelligence Suite for sifting through medical data. NICE uses machine-learning algorithms to look at published medical research, categorize it, and feed it to volunteer citizen scientists which then re-categorizes and processes it. This leaves researchers time to go through the final data, interpret and understand it, without having to waste time on the way. It also forms a virtuous cycle, whereby the citizen scientists feed the computer algorithm data and improve it, and the computer algorithm feeds the volunteers better data, speeding up their work.

Medical professionals need to be aware of current trends and how medical research is progressing, but the shear amount of papers and information available is an impossible feat to control.  Cortana can smartly parry down the data and transform it into digestible, useful material.

Whitney Grace, April 17, 2017

  • Archives

  • Recent Posts

  • Meta