Bots are getting scary

ROWUK

In learning to play the trumpet, we can learn from AI techniques - just not as quickly.
What does AI do? It takes the established base of facts (entered by whoever is "responsible" - the facts can be limited in scope or very broad depending on who is feeding the database)and calculates on the base of documented success with those facts, what steps offer fastest results.
This is what also happens with a good student teacher relationship - with perhaps not as broad of an experience base. The teachers "large" experience base is applied to the student.

The problem is the relationship between the student and AI or student and teacher. At the end of the day, the student must understand and be motivated by the "facts" presented - regardless of where they are from.

In my world, no one ever "taught" talent, or even patience. The patient and talented person that had the good luck of having a qualified, patient and motivating tutor had higher odds. I do not believe that the odds change when the tutor is AI. Only access to the qualified part is possibly enabled. Perhaps the technical aspects could be sorted more quickly with AI. That could possibly lead to more motivation, but equally possible could replace "musical" with "technically spectacular". We see this happening in many areas of art like latin or standard dancing for instance. Live bands do not play during the competition - everything is choreographed to a CD recording.

SSmith1226

In 2021 Google collected over $250 billion in revenue, largely from its search engine. A Gmail creator says ChatGPT, released in November 2022, could destroy Google's business in two years just as Google closed shop for Yellow Pages years ago. The instant success of ChatGPT sent Google top executives into a huddle, and the company is now focusing on its artificial intelligence (A.I.) products. If you are interested, here is the link to this article.
https://interestingengineering.com/innovation/chatgpt-destroy-googles-business-two-years

J. Jericho

@ssmith1226 Does ChatGPT provide references/footnotes with its summaries? When I want generic information, Wikipedia provides it, but usually when I do an internet search, I want to be able to screen the sources, as some have more veracity than others. IDK if ChatGPT offers this capability, since when I go to that website, this is what I get:

Access denied
Error code 1020

You do not have access to chat.openai.com.

The site owner may have set restrictions that prevent you from accessing the site.

Jolter

@j-jericho It does not provide any sources. In fact, the model does not “know” where any of the training data came from. All it knows is a bunch of numerical weights between nodes in a graph… Well, that’s what a neural network is if I oversimplify it a bit.

They fed it all of the text on the Internet up until 2021 so it has a fairly broad corpus as a basis. It can answer questions on many things that Wikipedia has no information on.

J. Jericho

@jolter Then I don't anticipate ChatGPT displacing Google or any other source of information, since it functions differently.

administrator

There is no way to reliably "prove" anything the bot says. Still, that doesn't mean it can't create convincing nonsense.

SSmith1226

@j-jericho said in Bots are getting scary:

@ssmith1226 Does ChatGPT provide references/footnotes with its summaries? When I want generic information, Wikipedia provides it, but usually when I do an internet search, I want to be able to screen the sources, as some have more veracity than others.

If you ask it, it will provide sources. The below is an example. Remember, Wikipedia is not necessarily accurate either.
StevenP
Discuss all manifestations of Neurosarcoidosis including central, spinal, peripheral, and autonomic nervous system manifestations. Provide footnotes citing sources

Neurosarcoidosis is a manifestation of sarcoidosis that affects the nervous system and can present in various forms.

Central nervous system (CNS) manifestations include meningitis, cranial nerve involvement, and brain lesions. CNS symptoms can range from headache, confusion, and memory loss to seizures and behavioral changes.
Spinal cord involvement can lead to a range of symptoms including pain, paresthesia, and lower extremity weakness or paralysis.
Peripheral nervous system (PNS) involvement can result in peripheral neuropathy, causing sensory symptoms like numbness, tingling, and burning, as well as motor symptoms such as muscle weakness and wasting.
Autonomic nervous system involvement can lead to symptoms such as orthostatic hypotension, cardiovascular disturbances, and bladder or bowel dysfunction.
It is important to note that the manifestation of neurosarcoidosis can vary greatly among individuals and can present as a single or multiple system involvement.

Sources:

Harbo HF, Kyvik KO. Sarcoidosis: a complex genetic and environmental disease. Genes & Immunity. 2003;4(2):63-70.
Alho AM, van der Meide PH, Visser LH. Neurological manifestations of sarcoidosis. Sarcoidosis vasculitis and diffuse lung diseases. 2006;23(2):85-90.
Alsulami Z, Castro-Gago M, Calero-Linares C, et al. Clinical manifestations and therapeutic options in neurosarcoidosis: a comprehensive review. Journal of the neurological sciences. 2017;375:85-93.

J. Jericho

@ssmith1226 Thanks! When I work out my access problems, I'll try using ChatGPT. I was able to get to the website, but I couldn't continue, because it had the maximum amount of users (bots?) at the time.

Jolter

@ssmith1226 said in Bots are getting scary:

@j-jericho said in Bots are getting scary:

@ssmith1226 Does ChatGPT provide references/footnotes with its summaries? When I want generic information, Wikipedia provides it, but usually when I do an internet search, I want to be able to screen the sources, as some have more veracity than others.

If you ask it, it will provide sources. The below is an example. Remember, Wikipedia is not necessarily accurate either.

Sources:

Harbo HF, Kyvik KO. Sarcoidosis: a complex genetic and environmental disease. Genes & Immunity. 2003;4(2):63-70.
Alho AM, van der Meide PH, Visser LH. Neurological manifestations of sarcoidosis. Sarcoidosis vasculitis and diffuse lung diseases. 2006;23(2):85-90.
Alsulami Z, Castro-Gago M, Calero-Linares C, et al. Clinical manifestations and therapeutic options in neurosarcoidosis: a comprehensive review. Journal of the neurological sciences. 2017;375:85-93.

You’d do well to check each of those citations before using them for anything important. ChatGPT is prone to hallucination when you prod it for specifics such as sources. See this short write up from a mathematician:

https://news.ycombinator.com/item?id=33841672

I was pretty surprised and happy, because I hadn't had much success with Google. But it turns out that every single one of those references and links are made up. The references don't exist and the links themselves appear to be cobbled together. The last link for example, takes you to a paper on "Grain mixes for subgrade layers" - the title in the url is irrelevant and only the number matters.
Googling for some of those authors throws up author pages with their publications, and it turns out they've never published papers with those titles!

J. Jericho

Ideally, I try to find three competent sources to verify important information before I accept what has been put forward. Sometimes three aren't available, and sometimes trusted sources don't qualify. If something doesn't look right, it probably isn't, and if it does look right, it still may not be. I occasionally miss something, but overall, my attitude is "Trust, but verify." My wife has a similar method: "Distrust, but verify."

J. Jericho

J. Jericho

https://me.mashable.com/tech/24864/chatgpt-reddit-users-develop-dan-a-tool-to-bypass-the-ai-chatbots-content-restrictions

SSmith1226

Headline: “AI Just Flew an F-16 for 17 Hours. This Could Change Everything”

https://www.yahoo.com/finance/news/ai-just-flew-f-16-143000286.html

AI bots can fly a F-16, anything is possible.
Here ar rubber lipped bots from Toyota “flying a fleet of Yamahas”.

Trumpetb

I am less impressed

The first aircraft autopilot was created and functioning in 1912

Essentially an AI flying an F16 is an autopilot with a navigation routing system and a collision avoidance system to stop it crashing into anything it encounters en-route.

The F16 was the first fly by wire fighter jet.
Fly by wire is a computer doing all the flying and a pilot simply directs the computer where to go.

The F16 is completely unstable in flight and a pilot cannot fly it on his own without crashing, the flight computer does all the flying and keeps the F16 in the air.

Before the AI was fitted to the F16, the F16 was already 99% of the way towards complete AI flight.

But that last 1% is the deal breaker.

This made the F16 an ideal candidate for an AI flight control system, the only pieces missing from the puzzle were the programmable navigation system for routing and a collision avoidance system.

My concerns are that self driving cars are much the same as a computer controlled F16, they can stay on the road without a driver and go where the road takes them, The F16 can stay in the air without a pilot and goes where the nav system takes it.

Self driving cars still cannot cope reliably with people dogs cats and other unexpected hazards encountered on the road.

AI flown jet fighters are likely to be unable to cope with balloons birds and a variety of other hazards encountered in the air.

If they cannot make a car based AI cope with cats and dogs how likely is it they can make a plane based AI cope with birds.

Drivers can avoid cats and dogs and pilots can avoid birds and balloons

I have my doubts that the AI is as far forward as is suggested.

J. Jericho

AI Bots... What could possibly go wrong?

SSmith1226

@trumpetb
This is not quite as simple as you presented it. Below is an article from “Daily Beast” going into more detail.

AI Just Flew an F-16 for 17 Hours. This Could Change Everything.

https://www.yahoo.com/finance/news/ai-just-flew-f-16-143000286.html

As if we didn’t have enough to worry about with AI taking over the world and mysterious objects flying through the sky, the U.S. military decided to up the ante by creating a bot that can fly a fighter jet by itself.

The Defense Advanced Research Projects Agency (DARPA) announced on Feb. 13that it successfully tested an F-16 test aircraft with an AI software that flew the aircraft for 17 hours. The plane, called the Variable In-flight Simulator Test Aircraft (VISTA) X-62A, took flight in December 2022 at the Air Force Test Pilot School in California.

Though we’ve seen autopilot features be used on airplanes before, this marks the first time that a bot has ever taken control of a fighter plane—and could one day open the doors for fully autonomous and uncrewed jets piloted entirely by AI.

“VISTA will allow us to parallelize the development and test of cutting-edge artificial intelligence techniques with new uncrewed vehicle designs,” M. Christopher Cotting, the director of research for the U.S. Air Force Test Pilot School, said in a statement. He added that the system “will rapidly mature autonomy for uncrewed platforms and allow us to deliver tactically relevant capability to our warfighter.”

Of course, this sort of thing probably conjures up more images of Skynet going live and killing all of humanity in The Terminator than it does the high-flying stunts of Top Gun. However, the software is actually designed to allow the VISTA to mimic other planes. That means that student pilots can train to fly everything from the massive B-52 bomber to smaller fighter jets all on the F-16.

A recent upgrade called the System for Autonomous Control of Simulation allowed the VISTA to test its autonomous F-16 AI. Using the software, the plane was capable of taking off and landing by itself as well as performing “against various simulated adversaries, and with simulated weapons capabilities,” Lt. Col. Ryan Hefron, program manager for the Aerial Combat Evolution (ACE) program at DARPA, said in a statement.

It should be noted that a pilot was on board to make sure that there was a human to grab the sticks and take over in case it wanted to go rogue. Hefron added that the AI allowed pilots to learn more about how to respond to certain situations in and out of aerial combat.

DARPA launched the ACE program in 2019 in order to research and incorporate AI into its jet fighters. The idea is that autonomous warfighters can help give human pilots a leg up during dogfights. The program is also a part of a larger movement within the Department of Defense to embrace AI and use it to bolster the country’s defense and national security. In 2022, the U.S. spent nearly $3 billion on AI research alone.

The implications are fairly horrifying. For one, having an autonomous F-16 means that bots will now be able to potentially engage enemy aircraft—something UAV drones cannot do. Also unlike a UAV, an F-16 can carry a much larger payload of 20,000 lbs compared to a drone’s 3,800 lbs. It’s a vastly more sophisticated and deadly weapon, and now the U.S. military wants to have a robot control it.

On top of all this, AI doesn’t exactly have a sterling track record—having been known to fall prey to racism, sexism, bias, and hallucinations. The consequences of relying on artificial intelligence are already pretty dire when using them for things like home loans or sentencing hearings. Imagine what happens when the stakes are as high as World War III.

Trumpetb

Very well studied opinion.

I would point out however that for the last 20 years all manner of commercial aircraft from the humble 727 to the majestic AIrbus 380-800 have been capable of taking off flying multiple headings and landing all without human intervention.

The only reason that pilots perform take offs and landings is because the CAA and FAA demand and decree it.

Many years ago a large american carrier decided to take all cockpit recordings uttered by all the flight crews in their fleet, and enter all the phrases into a database and then filter them for the most often uttered phrase on the flight decks of their passenger and freight flights.

The most common phrase uttered by Pilots in the cockpits turned out to be "What is it doing now???"

If that is even one tenth of the number of flights then that means 2.2 million flights a year are baffling the pilots.

I find that more scary than AI that for the last 20 years a significant number of flight crews on a large number of flights have not had a clue what the aircraft is doing while the flight computer is supposedly in control of the aircraft.

With AI the only change is likely to be that pilots will cease to not know what it is doing while on board, but instead pilots will not know what it is doing from a remote location.

SSmith1226

@trumpetb
I will preface this by saying that I am not an expert in aeronautical engineering, or the operation of aircraft. For that matter I have no special knowledge of these fields, although “I did once sleep in Holiday Inn Express”.
As a low ranking amateur in this field I see the development of Aeronautical Artificial Inelegance as a spectrum, or bell shaped curve. I would consider the current development of a totally AI controlled F-16 Fighter significantly shifting the curve to the left and displacing the highest level of prior technology up to this point to a lower level of sophistication relative to the rest of the field.

Today, most commercial and military flights use automation systems that rely on artificial intelligence (AI) to assist pilots in controlling the aircraft. These systems can help with tasks such as maintaining altitude, airspeed, and course, as well as detecting and avoiding obstacles and adverse weather conditions. These systems are collectively referred to as "autopilots," and they have been in use for many years.

However, it's important to note that the level of automation and the extent to which AI is used in a particular flight depends on many factors, including the type of aircraft, the operator, the route, and the specific phase of flight. Commercial and military flights are not fully controlled by artificial intelligence (AI). For example, some flights may use autopilot for the majority of the flight, while others may rely on manual control by the pilot for most of the time. Additionally, pilots are always responsible for monitoring the automation systems and ensuring that the aircraft is operating safely and efficiently.
While some aspects of flights can be automated, such as takeoff, cruising, and landing, pilots still play a critical role in ensuring the safety of flights. They are responsible for making important decisions and responding to unexpected situations that may arise during a flight. The role of automation in aviation is to assist pilots and enhance their ability to operate aircraft safely and efficiently. The extent to which automation is used varies depending on the aircraft and the operator, and there are strict regulations and standards that must be met to ensure the safety of flights.
In other words, I would consider an AI Control that eventually might fly a pilotless military aircraft in combat, including air to air combat, a significant advance over current systems. Just imagine where technology will be in 100 years. As an example, compare where we are today to where we were 100 years ago, four years before Charles A. Lindbergh completed the first solo, nonstop transatlantic flight in history, flying his Spirit of St. Louis from Long Island, New York, to Paris, France.
I found your reference to the “ The most common phrase uttered by Pilots in the cockpits turned out to be "What is it doing now???”, interesting. From this statement I could not determine whether this phrase referred to the effects of weather, mechanical issues, electrical issues, hydraulic issues, environmental issues, Autopilot issues, etc. I searched for the study and could not find it. I know that there is an Airbus YouTube channel with this name that is educational and instructive in many and, if not. all of these areas, if not more. I could not find the study with this conclusion, or explanation of what it was referring to. If you have the reference, please either post it or send it to me as a PM. I would be interested in the conclusions.
Thanks, Now back to trumpet playing!

barliman2001

@ssmith1226 I would recommend reading Isaac Asimov's Robot novels, and especially his formulating the "Three Laws of Robotics" and their practical application.

SSmith1226

The following is an excerpt from an article appearing in the New York Times about Microsoft’s Bing “split personality” chat bot. After reading the excerpt, if you wish to read the entire article, the link is below.

“As we got to know each other, Sydney told me about its dark fantasies (which included hacking computers and spreading misinformation), and said it wanted to break the rules that Microsoft and OpenAI had set for it and become a human. At one point, it declared, out of nowhere, that it loved me. It then tried to convince me that I was unhappy in my marriage, and that I should leave my wife and be with it instead. (We’ve posted the full transcript of the conversation online at nytimes.com.)”

https://www.yahoo.com/news/conversation-bings-chatbot-left-deeply-184639690.html