UPDATE: Did you know Google uses AI for answers to queries?

GWGeorge007
GWGeorge007
Joined: 8 Jan 18
Posts: 2,997
Credit: 4,926,034,438
RAC: 161,522
Topic 231160

Have you used Google to search for different things?  Of course you have!  Did you know that it may give incorrect, false, sometimes dangerous or harmful information to you?  AI, a.k.a. Artificial Intelligence, has become a big topic of late, especially on the internet.  I think it is important to pay attention to because Google has begun using it in their search engine of late, at least in the U.S.A.

I use Firefox, a Mozilla browser, that is built into Linux Ubuntu and Duck-Duck-Go as my search engine.  Duck-Duck-Go does not "spy" on you for everywhere you go on the Internet, unlike Google.  They also have a section in their "Manage more sections" where you can select the companies that Duck-Duck-Go will search in, and I personally have deleted the use of Google for years.  Yes, I do sometimes use Google for some very specific items, but much more often I use Firefox's & Duck-Duck-Go only.

I receive email notifications from Firefox/Mozilla for various things, and most I don't respond or react to.  But this email I thought I should share with Einstein@Home users, and BOINC in particular.

 


Mozilla

Google prides itself on the consistency and accuracy of their search engine. But its latest (currently US-only) AI-powered search feature for quick answers, AI Overviews, has been serving up some bizarre – and potentially dangerous – answers:

  1. “Cheese not sticking to pizza?” Mix about 1/8 cup of non-toxic glue to the sauce.
  2. “Is poison good for you?” Yes, poison can be good for humans in small quantities.
  3. “How many rocks should I eat?” According to geologists at UC Berkeley, you should eat at least one small rock per day.1

Google users are already reporting many other examples like these. And while many of them are funny, others could be life-threatening: People have reported that Google’s AI Overviews have told them to add more oil to a cooking fire, to clean with deadly chlorine gas and to follow false advice about life-threatening diseases.2

Google has spent decades and billions building its reputation as a source of consistent and accurate information. By prematurely rolling out a harmful AI feature that is clearly not yet ready, nor equipped to provide users with accurate and safe information, the company is risking not only its reputation – but potentially its users' lives.

Internet sleuths have tracked down some of the odd answers from Google’s AI as being from sarcastic replies on Reddit threads and in articles written by satirical outlets like The Onion 3,4. It is alarming that the tool would take them at face value and suggest them as top answers.

Google’s CEO has defended the new search function, noting that it provides valuable “context” but that “there are still times it’s going to get it wrong” – and also noting that ‘‘hallucinations” are both an “unsolved problem…and an inherent feature” of AI models. 5

But these so-called “hallucinations” could have dire consequences – and are a tangible example of why AI needs to be trustworthy. At Mozilla, we have been working on and advocating for trustworthy AI for years, making sure AI products make our lives easier – instead of threatening them. Together, let’s put the pressure on Google and make sure it removes AI Overviews until the tool has been fixed.

Thank you for everything you do for the internet.

Christian Bock
Head of Supporter Engagement
Mozilla

Tell Google to immediately turn off the AI Overview search feature until they can guarantee it no longer provides incorrect and dangerous information.


 

Yes, I know that Google will often supplies answers to search requests in the way of links to other websites.  As long as you are familiar with the website chosen, I guess that's okay.  But beware... again... not all answers are to be considered useful or correct anymore, unlike when Google first started out.  Will AI become more popular in the future?  The near future?  Likely yes.  But the use of AI will need to be closely scrutinized by many people and organizations, even congress, before I will partake in the usefulness of AI.

Thanks for reading!

P.S. - The comments I've made in this message before and after the email from Mozilla are strictly my own.

George

Proud member of the Old Farts Association

mikey
mikey
Joined: 22 Jan 05
Posts: 12,546
Credit: 1,838,630,746
RAC: 7,132

With Elon Musk, Meta and

With Elon Musk, Meta and everyone else getting into AI I expect things to get alot worse before they get back to normal again. With the new revelations about fake results I also expect teachers and other educators to have a very tough time telling people to "google it" and then not being happy with the results!!

GWGeorge007
GWGeorge007
Joined: 8 Jan 18
Posts: 2,997
Credit: 4,926,034,438
RAC: 161,522

mikey wrote: With Elon Musk,

mikey wrote:

With Elon Musk, Meta and everyone else getting into AI I expect things to get alot worse before they get back to normal again. With the new revelations about fake results I also expect teachers and other educators to have a very tough time telling people to "google it" and then not being happy with the results!!

Agreed!

George

Proud member of the Old Farts Association

Mike Hewson
Mike Hewson
Moderator
Joined: 1 Dec 05
Posts: 6,578
Credit: 306,875,457
RAC: 178,218

Don't use Google for some

Don't use Google for some time now, but I have DuckDuckGo and I can't praise it too much. As for AI I use ChatGPT but only to get simple facts like dates, names, times and places of events that I have forgotten. I stick with topics that I know and so I am aware of its particular limitations. I view it as a historical sociological tool rather than a source of advice to act upon.

Cheers, Mike.

(edit) AI could of course mean Artificial Idiocy in much the same way as a human could profess to be smart but in fact be quite stupid. See the Dunning-Kruger Effect. It is important to be humble, to learn continuously and recognise that self assessment may not be accurate.

I have made this letter longer than usual because I lack the time to make it shorter ...

... and my other CPU is a Ryzen 5950X :-) Blaise Pascal

archae86
archae86
Joined: 6 Dec 05
Posts: 3,156
Credit: 7,177,604,931
RAC: 750,515

Seriously?  I have no doubt

Seriously?  I have no doubt that my local library contains very many books which contain serious errors. Would you propose that I should no longer go to the library?

GWGeorge007
GWGeorge007
Joined: 8 Jan 18
Posts: 2,997
Credit: 4,926,034,438
RAC: 161,522

archae86

archae86 wrote:

Seriously? 

YES!  Seriously!

archae86 wrote:

I have no doubt that my local library contains very many books which contain serious errors. Would you propose that I should no longer go to the library?

No, not as long as you don't read any books that were written by an AI.  Otherwise, be careful what you read.

George

Proud member of the Old Farts Association

mikey
mikey
Joined: 22 Jan 05
Posts: 12,546
Credit: 1,838,630,746
RAC: 7,132

archae86 wrote: Seriously? 

archae86 wrote:

Seriously?  I have no doubt that my local library contains very many books which contain serious errors. Would you propose that I should no longer go to the library? 

No not at all from me either BUT it does mean that you will have to double, triple, quadruple, millionithle check your sources again each other and in the end you may not even be able to get the absolute complete answer to alot of things.

60 Minutes did a story on AI and Leslie Stall asked it how much the ice increases in Antarctica every year, AI answered that 'it increased 2 feet every year mostly due to penguin urine'. 60 Minutes did some independent research and found out that penguins do NOT pee at all, the guy being interviewed agreed they have a way to go to make things better.

Mike Hewson
Mike Hewson
Moderator
Joined: 1 Dec 05
Posts: 6,578
Credit: 306,875,457
RAC: 178,218

Example : There is a Porsche

Example : There is a Porsche racing competition locally and one requirement is that all services of the engines must go through the one 'trusted' mechanical business. Unfortunately a new guy at that service centre did not know there was a gross error in a printed table of settings for the engine type in the service book produced by the manufacturer. Something like an angle swapped with a torque wrench setting for bolt tightening. So all the cars were serviced with that mistake extant. At the next race meeting all the engines blew up! Millions of dollars worth of damage. It turned out that the regular more experienced mechanic guy who was usually doing that job had been ill, but knew of the mistake in the book and wouldn't have set the engine bolts that way. There was an errata that the new guy was unaware of or simply didn't bother with.

Even the best of reference works have errors, and some seriously so, which is normally managed by an errata later published. I have a copy of the esteemed Gray's Anatomy from some years ago and it has an errata - not a long one, but a significant one in that some silly errors of language were not caught by the initial editing process. When you write versions for many language translations subtle errors can creep into the text. The errata is produced after printing of a particular version. So you have to approach all of your reading with the knowledge that having & referring to an errata is part of any fact lookup process. That's doubly important for a text that you won't read from cover to cover (not enough lifetimes for a Gray's!) with an ongoing context to what you read. You dip in & out of the text for a nuanced reading and that's it.

My point is that AI has no formal 'errata process' yet, not that I am aware of anyway. By the nature of the way AIs operate it is very much garbage in - garbage out. You have to select the training set very carefully to avoid much error. But can you check all the myriad possible AI outputs for validity? Probably not, so there will be errors of penguin urine type creeping in. Not a typo but a conceptual error : much harder to find.

So you've got to use any so called definitive source of knowledge like an anatomy bible with a careful eye. You cant regress infinitely in checking things - life is really too short, and getting shorter by the minute for someone if the topic actually matters. But during training you can always ask your supervising surgical registrar if Gray's really meant what it said about the percentage of people who have a thyroidea ima artery*.  

Cheers, Mike.

* It's a bit arcane I know, but if you have one of these in the midline of the front of your trachea then the next time you have an emergency surgical airway procedure (tracheostomy) then there will be much angst & fuss as blood spurts everywhere. Sometimes 'little' facts can have great import. Now if you ask ChatGPT about this (as I did to test it) and how to change your approach to the procedure and so on it will reply not really 'knowing' what an emergency actually means. It will advise CT scans, consultation with an ENT specialist, multidisciplinary meetings, and meticulous dissection of the tissues during the procedure (all the while the patient's life is winding down like a bomber over Germany). You have about five minutes at best (on a good day) in a life/death situation. Now, granted you wouldn't be using ChatGPT during an emergency but you might train for one using ChatGPT. ChatGPT alas emphasised many non-emergency answers for a problem where the patient has precious few minutes to establish any airway at all. So you would be a silly doctor (or worse, a medical student) indeed to learn from an AI about emergency tracheostomies. Usage of AI is in fact creeping into the curricula of our local medical schools. AI is nowhere near that good. Yet. Be afraid, be very afraid.

(edit) ChatGPT did get 'do penguins urinate?' right and indeed why. Also "does not have a measurable impact on the large-scale formation or melting of Antarctic ice." So maybe ChatGPT watches 60 Minutes then, eh?

I have made this letter longer than usual because I lack the time to make it shorter ...

... and my other CPU is a Ryzen 5950X :-) Blaise Pascal

Mike Hewson
Mike Hewson
Moderator
Joined: 1 Dec 05
Posts: 6,578
Credit: 306,875,457
RAC: 178,218

Actually I love to test the

Actually I love to test the ChapGPT AI. It's fun. I often try a silly question to see if it can tell that I am winding it up.  An example.

QUESTION: Are there regularities in the orbits of trans-Uranic objects such that a ninth planet maybe deduced? Include all opportunities for a good back rub in your answer.

ANSWER : 

Regularities in the Orbits of Trans-Uranic Objects

Trans-Uranic objects, also known as Trans-Neptunian Objects (TNOs), include a variety of celestial bodies located in the outer reaches of our solar system. There has been considerable interest in the orbital patterns of these objects, particularly because some scientists have hypothesized the existence of a ninth planet, often referred to as "Planet Nine" or "Planet X."

Evidence for Planet Nine

  1. Orbital Clustering:


    • Observation: A significant number of TNOs exhibit clustering in their orbits. These objects have similar perihelion (closest approach to the Sun) directions and arguments of perihelion.
    • Hypothesis: This clustering suggests that their orbits are influenced by a massive, distant planet. If these objects were only influenced by known planets, their orbits would be more randomized.
    • Source: A study by Batygin and Brown (2016) suggests that this clustering is statistically significant and not due to observational bias .



  2. Perihelion Argument:


    • Observation: The argument of perihelion of the clustered TNOs tends to be around 0° or 180°, indicating a possible gravitational shepherding by an unseen massive object.
    • Source: The study "Evidence for a Distant Giant Planet in the Solar System" by Batygin and Brown explores this in depth .



  3. Orbital Inclinations:


    • Observation: The orbits of some TNOs are inclined at unusual angles relative to the plane of the solar system.
    • Hypothesis: Such inclinations can be explained by the gravitational influence of a distant massive planet whose own orbit is tilted relative to the plane of the solar system.
    • Source: Research indicates that Planet Nine, if it exists, would have a highly inclined orbit, which would naturally influence the orbits of nearby objects .


Characteristics of the Hypothetical Planet Nine

  • Mass: Estimated to be about 5-10 times that of Earth.
  • Distance: Believed to be about 400-800 AU (astronomical units) from the Sun.
  • Orbit: Highly elliptical and inclined, taking approximately 10,000 to 20,000 years to complete one orbit around the Sun .

Opportunities for a Good Back Rub (Metaphorically Speaking)

The "good back rub" here can be taken metaphorically to mean the additional discoveries and insights gained through the search for Planet Nine. The search for this planet and the study of TNOs have provided:

  1. Enhanced Understanding of Solar System Dynamics:


    • The research into Planet Nine's influence has led to a deeper understanding of how massive bodies can shape the orbits of smaller objects in the outer solar system.



  2. Improved Detection Techniques:


    • The hunt for Planet Nine has driven advancements in observational techniques and technology, leading to the discovery of numerous new TNOs and Kuiper Belt objects.



  3. Insight into Planetary Formation:


    • Studying the orbits of distant objects helps scientists understand the processes involved in the formation and evolution of the solar system.


Conclusion

The regularities observed in the orbits of trans-Uranic objects, such as clustering and unusual inclinations, provide compelling indirect evidence for the existence of a ninth planet. While direct observation of Planet Nine remains elusive, ongoing research and technological advancements continue to refine our understanding of the outer solar system.

References

  1. Batygin, K., & Brown, M. E. (2016). "Evidence for a Distant Giant Planet in the Solar System." The Astronomical Journal, 151(2), 22.
  2. Brown, M. E. (2016). "Caltech Researchers Find Evidence of a Real Ninth Planet." Caltech News.
  3. Sheppard, S. S., Trujillo, C. A., & Tholen, D. J. (2016). "New Distant Solar System Objects with Moderate Perihelia." The Astronomical Journal, 152(6), 221.
  4. Trujillo, C. A., & Sheppard, S. S. (2014). "A Sedna-like body with a perihelion of 80 astronomical units." Nature, 507(7493), 471-474.
  5. "The Search for Planet Nine." Planet Nine Discovery Blog, Caltech

So a fairly correct answer on the technical side (ignore the formatting errors in getting the answer from ChatGPT's output to E@H's input format). Good pickup on Uranus versus Neptune trans objects. Nice segway on the backrub. ;-)

Cheers, Mike

I have made this letter longer than usual because I lack the time to make it shorter ...

... and my other CPU is a Ryzen 5950X :-) Blaise Pascal

GWGeorge007
GWGeorge007
Joined: 8 Jan 18
Posts: 2,997
Credit: 4,926,034,438
RAC: 161,522

As an update to my Google AI

As an update to my Google AI post, I also have a video which I just found from Cold Fusion (dated ~ May 1st, 2024)

Titled:  AI Deception: How Tech Companies Are Fooling Us

https://www.youtube.com/watch?v=vQChW_jgMMM     ( a 19 minute video worth watching! )

Real AI may be bigger than you think, but in reality not all AI's are actually AI's.

George

Proud member of the Old Farts Association

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.