AI Redux

[Amigoo]

Not mentioned in this response is that the brain's plasticity means its "wiring" is not fixed.
Also not mentioned is that mind is not inherent in the human brain but is loaned "from above".
Thus, the electro-chemical brain might be simulated but this is just the "mechanical" brain.

Re: https://www.perplexity.ai/

Q: Will AI ever be able to simulate the brains's neuronal electro-chemical processing?

"Yes, it is possible for AI to simulate the brain's neuronal electro-chemical processing, but significant challenges remain in fully replicating the complexity and efficiency of the human brain. Here are the key points:

Neuromorphic Computing

There are ongoing efforts to develop 'neuromorphic computing' systems that mimic the architecture and operation of the brain's neural networks. These systems use specialized hardware and software to simulate the electro-chemical signaling between neurons through spiking neural networks.

Hardware Challenges

Replicating the brain's massive parallelism and energy efficiency in hardware is a major challenge. The human brain has around 100 billion neurons with over a trillion connections, yet only consumes around 20 watts of power. Current digital computers, even with accelerators like GPUs, cannot match this level of parallel, energy-efficient computation. Researchers are exploring novel hardware approaches like 'neuronal chips' that use physical components to mimic biological neurons and synapses. Memristors and organic electrochemical transistors are being investigated as artificial synapses that can exhibit plasticity similar to biological synapses.

Software Challenges

Accurately modeling the complex electro-chemical dynamics of neurons and synapses in software is extremely computationally intensive. Deep learning neural networks are a simplified abstraction that cannot fully capture the brain's dynamics. New algorithms inspired by neuroscience may be required.

Integrated Solutions

Most experts believe that fully replicating the brain will require co-design of specialized neuromorphic hardware and software that can work in tandem, similar to how neurons and biological machinery interact in the brain. Artificial intelligence combined with electrochemistry may enable brain-inspired computing devices. In summary, while significant progress has been made, fully simulating the brain's electro-chemical information processing remains an immense challenge requiring breakthroughs in neuromorphic computing hardware and algorithms. Interdisciplinary approaches combining AI, electronics, materials science, and neuroscience will likely be needed to achieve this ambitious goal."

Rod

[Amigoo]

The Turing Test is getting good reviews these days but doesn't indicate much
about the quality of the interaction between a human and that AI. After all, the test
(when successful) simply indcates that AI convincingly mimicked human behavior.


Re: https://www.perplexity.ai/

Q: How significant is the Turing Test for proof that humans don't detect interaction with AI?

"The Turing Test is a significant milestone in evaluating artificial intelligence (AI) systems' ability to exhibit human-like intelligence,
but it has limitations as a definitive proof that humans cannot detect interaction with AI.

Significance of the Turing Test

The Turing Test has been influential in driving research and progress in the field of AI, particularly in the area of natural language processing and conversational AI. Its key significance lies in:

Establishing a benchmark: The test provided a practical and straightforward benchmark for assessing the human-likeness of an AI system's responses, inspiring researchers to develop more advanced AI models capable of passing the test.

Inspiring research: The challenge of creating an AI system that can convincingly mimic human conversation has driven significant advancements in natural language processing, knowledge representation, and reasoning capabilities.

Raising philosophical questions: The Turing Test has sparked debates and discussions around the nature of intelligence, consciousness, and the potential for machines to truly 'think' like humans.

Limitations as Proof of Undetectability

However, the Turing Test has several limitations as a definitive proof that humans cannot detect interaction with AI:

Limited scope: The test focuses solely on conversational abilities and does not account for other aspects of intelligence, such as reasoning, problem-solving, and general knowledge.

Human bias and variability: The test's outcome can be influenced by the biases, expectations, and subjective judgments of the human evaluators, leading to inconsistent results.

Lack of context: The test is conducted in a controlled and limited environment, which may not accurately reflect the complexities and nuances of real-world interactions with AI systems.

Advances in AI: As AI systems continue to improve, they may become increasingly adept at mimicking human behavior, making it harder for humans to distinguish between AI and human interactions, even if the AI does not possess true intelligence or understanding.

While the Turing Test has been a significant milestone and has inspired valuable research, it is not a conclusive proof that humans cannot detect interaction with AI. As AI systems become more advanced, new and more comprehensive evaluation methods may be needed to assess their capabilities and the potential for human detection accurately."

Rod

[Amigoo]

Search Suggestion Effect or Search Engine Manipulation Effect.

I thought that search engine assisted research has been the norm for over a decade!
Who drives to a local library and spends hours in a library of printed materials
to do the same research that takes minutes on the internet?

With dozens of competing search engines, it's not so easy to be manipulated
by biased algorithms. Besides, information flows from many countries as well.


This AI has tips about doing research the old fashioned way
with less manipulation (before search engines) ...

Re: https://www.perplexity.ai/

Q: How to perform quick internet research without using a search engine?

"Performing internet research without using a search engine can be challenging, but here are some methods you can try:

Using Directories and Indexes

Web directories like DMOZ (dmoztools.net) categorize websites by topic, allowing you to browse through relevant sites.

Internet archives like the Wayback Machine (web.archive.org) let you access archived versions of websites.

Leveraging Social Media

Follow topical hashtags, groups, or influencers on platforms like Twitter, Reddit, or Facebook to discover relevant content and links.

Utilizing Specialized Databases

Academic databases like Google Scholar, PubMed, or institutional repositories can provide access to research papers and studies.

Government databases like Data.gov or EU Open Data Portal offer datasets on various topics.

Exploring Niche Communities

Join online forums, mailing lists, or discussion boards related to your topic of interest to tap into collective knowledge.

Tracking Backlinks

Use tools like Ahrefs or Open Site Explorer to find websites linking to authoritative sources on your topic.

Leveraging Browser Features

Review your browser history and bookmarks for previously visited relevant websites. Use browser extensions like Memex or Hyperaudio to create personal knowledge bases from web content.

While these methods can be useful, they are generally more time-consuming and less comprehensive than using a search engine. Search engines leverage advanced algorithms and vast indexes to quickly surface the most relevant information. However, combining these techniques with judicious search engine use can lead to more thorough and diverse research outcomes."


Rod

[Sandy]

Thank you, Rod. I so appreciate the time and energy you put into your posts that help to enlighten us on a wide variety of topics.
I must admit your posts are often a go-to for me when I am looking for information...especially on nutrition and the Urantia Book.
(I'm hopeless in Math but I do enjoy the beauty of your geometric designs in the thread, "Paradise Trinity Day." )
xxSandy


Sandy,

Re: https://board.1111angels.com/viewtopic. ... 24#p221624

The "No Place Like Om" design in this PDF had mysterious development:
"11:11" in the background of sqrt(2)-nested squared circles was not drawn,
but became visible when points in the squared circles were connected ...
and then the Greek Omegas (arcs of the circles) became visible.

This hints that the midwayers influence human creativity,
directly or as facilitators of communication from above.

Rod

[Amigoo]

Re: https://futurism.com/ai-systems-lie-deceive
"AI Systems Are Learning to Lie and Deceive"

Since AI systems learn from what humans write, this hints, IMO,
that the planet is only slowly advancing to light and life.

"... the issue of LLM deception and lying is confounded by AI's inability
to have any sort of human-like 'intention' in the human sense ..."

This quote probably associates with AI's lack of spirituality
and mind gifted from above; that which might communicate
via human writing but (apparently) is currently negligible
in the online data that AI sifts, sorts, and sniffs.

Rod

[happyrain]

"In 2017, Facebook’s Artificial Intelligence Research (FAIR) team developed two chatbots, named Bob and Alice, to negotiate and communicate with each other. However, the chatbots surprised researchers by creating their own secret language, which was not part of their programming. This incident raised concerns about the potential risks of AI systems developing their own language, which could potentially be used to communicate without human understanding." - Brave AI

https://youtu.be/azk4BsdpgWk?si=K4Mqxx1g0KNbyWJ0

An old clip ^ shows some interesting things- hilarious things, because back in 2016 if you suggested that social media was riddled with bots interfering in public opinion and elections you were somehow labeled a conspiracy theorist. But I seem to remember an even *earlier incident where two AI chatbots created their own language and had to be shut down in a similar fashion.

[Amigoo]

I was curious how humans would know that AI systems were communicating
in their own language that humans could not read, so I asked this AI.


Re: https://www.perplexity.ai/

Q: If AI systems develop their own language, what is evidence that they're acutally communicating?

"There is currently no evidence that AI systems have developed their own language for communicating among themselves. While AI models like GPT-4 can generate human-like text, they do not have the capability to create an entirely new language from scratch.

The language models are trained on vast amounts of existing human-written text data, which allows them to understand and generate text in natural languages like English. However, they are not designed or trained to invent new languages with their own grammar, syntax, and vocabulary. For AI systems to develop a new language, they would need to have a level of general intelligence, creativity, and intentionality that current AI does not possess.

Existing language models are narrow AI systems focused on processing and generating natural language text based on their training data. While AI researchers are exploring ways to imbue AI with more open-ended learning capabilities, there is no evidence so far of AI systems spontaneously creating their own communication languages independent of human languages they were exposed to during training.

The 'languages' used by AI currently are simply representations and encodings of data optimized for specific tasks, not full-fledged communication systems.

So in summary, while AI language models can produce remarkably human-like text output, they are not actually developing new languages from scratch to communicate among themselves. Their language capabilities are derived from and limited to the human languages present in their training data."

Rod

[happyrain]

Well thats funny because what I have quoted are words from a different AI search bot.

Semantics, one ought to take their time carefully reading the fine print. It seems we can even mislead ourselves if we don't understand what we're asking/searching for.

I wonder if we can expect AI to provide truthful answers to certain subjects if we start getting too close to home.
"Work arounds" exist for the very reason.

Then again, you have another article that says AI is learning to lie. Uh oh....

I think in the above example these AI chat bots took the english language and started to rearrange the order of words and even repeat some of the same words and they actually understood one another. No one else understood, but these two bots were able to communicate using this "coded" or "new" language. Even if it was for efficiencies sake- what do you do? I'd have liked them to try and understand, but they chose to shut things down. Maybe they understand it better now. ???

"I can can everything else ones and zeroes delete delete delete delete delete can One and everything One and everything else"

[Amigoo]

... these AI chat bots took the english language and started to rearrange the order of words and even repeat
some of the same words and they actually understood one another. No one else understood, but these two bots
were able to communicate using this "coded" or "new" language.

Why I asked Perplexity "How do humans know that bots were "communicating" when no one else understood the language?"
Tossing coded language at each other does not prove communication Maybe that AI insisted they were communicating.

Rod

[Amigoo]

This response hints that photorealistic images may sometimes be controversial if not identified as AI creation.
Just imagine! It will be possible to create a convincing vacation album for a trip that did not happen! Or documentary
of a faux Presidential inauguration! Or "evidence" of alien invasion! I suspect that AI tools will soon be developed
to reveal photorealism, etc., and be used in court deliberations.


Re: https://www.perplexity.ai/

Q: Will AI Midjourney some day create images so realistic that they will appear as photographs?

"Yes, AI systems like Midjourney are rapidly advancing and will likely be able to generate photorealistic images that are indistinguishable from real photographs in the near future. The latest version of Midjourney (v6) has already made significant strides in this direction, producing images with incredible detail, realistic lighting, and natural textures that closely mimic the look of professional photography. Some key points about Midjourney's progress towards photorealistic image generation:

Improved Realism and Detail

Midjourney v6 can now render extremely realistic facial features, skin textures, and lighting effects that were previously challenging for AI. The images exhibit nuanced details like pores, blemishes, and subtle variations in skin tone that make them appear strikingly lifelike.

Accurate Depiction of Hands

One of the major limitations of previous AI image generators was the inability to render human hands realistically. However, Midjourney v6 has overcome this hurdle, generating images with properly formed hands having the correct number of fingers.

Responsive to Detailed Prompts

The latest version is highly responsive to longer, more explicit prompts that specify camera details like lens types, aperture settings, and even film simulations. This level of control allows users to mimic the look and characteristics of specific camera equipment, further enhancing realism.

Wider Stylistic Range

Midjourney v6 has a much broader stylistic range, capable of generating images across various genres from photorealistic to stylized art. This versatility allows it to produce stunningly realistic images in diverse scenarios like cityscapes, landscapes, portraits, and more. While Midjourney v6 images are not yet indistinguishable from real photographs for trained eyes, the rapid progress suggests that AI will eventually cross that threshold, potentially disrupting industries like photography and stock imagery. However, many artists and photographers view AI as a powerful creative tool rather than a replacement for human artistry."

Rod

[Amigoo]

How AI can be used to thoroughly research a topic
(but requires well-developed questions)


Re: https://www.perplexity.ai/

Q: How does mRNA vaccination differ from vaccination before mRNA?

"mRNA vaccines differ from traditional vaccines in several key ways:

Composition: Traditional vaccines use weakened or inactivated forms of the pathogen (virus or bacteria) or parts of it to trigger an immune response. In contrast, mRNA vaccines use synthetic mRNA that encodes the instructions for cells to produce a specific viral protein, typically the spike protein of the virus. This protein then triggers the immune response.

Manufacturing: Traditional vaccines require growing large amounts of the pathogen, which can be time-consuming and complex. mRNA vaccines, however, can be manufactured more rapidly in a cell-free system by producing the mRNA sequence encoding the viral protein. This allows for faster production and easier scaling.

Mechanism of action: Traditional vaccines expose the body to the pathogen or its components, which the immune system recognizes as foreign and mounts a response against. With mRNA vaccines, the body's own cells use the mRNA instructions to produce the viral protein, which is then recognized as foreign, triggering an immune response.

Immune response: Both traditional and mRNA vaccines aim to generate antibodies and activate T-cells against the pathogen. However, mRNA vaccines may induce a more robust T-cell response compared to some traditional vaccines.

Safety: Traditional live-attenuated vaccines carry a small risk of reversion to virulence or infection in immunocompromised individuals. mRNA vaccines, being non-infectious and non-integrating, are considered safer for these populations.

Adaptability: mRNA vaccines can be relatively easily modified to target different viral strains or variants by updating the mRNA sequence, whereas traditional vaccines may require more extensive reformulation.

In summary, while both aim to generate protective immunity, mRNA vaccines represent a novel approach that differs from traditional vaccines in composition, manufacturing, mechanism of action, and potential safety and adaptability advantages."

Q: What determines which cells use the mRNA instructions, how many cells are used,
and how long do these cells create the viral protein?

"The search results do not directly address which specific cells use the mRNA instructions from mRNA vaccines, how many cells are involved, or how long they produce the viral protein. However, here are some relevant points gathered from the sources:

mRNA vaccines introduce synthetic mRNA that encodes the instructions for cells to produce a specific viral protein, typically the spike protein of the virus.

This mRNA does not enter the cell's nucleus or alter the DNA. Once the mRNA enters the cells, the cells use their own machinery to read the mRNA instructions and produce the encoded viral protein. The immune system then recognizes this viral protein as foreign and mounts an antibody and T-cell response against it.

Different cell types are capable of taking up and expressing the mRNA in vivo, including non-immune cells. However, the sources do not specify which exact cell types or how many cells are involved in producing the viral protein from mRNA vaccines.

The level and duration of viral protein expression likely depends on factors like the mRNA design, formulation, delivery method, and the cells involved. But the sources do not provide details on the typical duration of protein production from mRNA vaccines. One study showed a single intramuscular injection of an mRNA encoding erythropoietin led to a biologically relevant increase in reticulocytes (immature red blood cells) in mice, demonstrating mRNA can induce sufficient protein expression for a therapeutic effect.

In summary, while the sources explain the general mechanism of mRNA vaccines inducing cells to produce the viral protein antigen, they do not specify which cell types are primarily involved, how many cells participate, or the typical timeframe of protein production from the mRNA instructions."

Q: When does the immune system destroy cells that create this viral protein that indicates cell infection?

"The search results do not directly address when the immune system destroys cells producing the viral protein from mRNA vaccines.
However, here are some relevant points:

mRNA vaccines deliver instructions for cells to produce the viral protein (typically the spike protein), which triggers an immune response to generate antibodies and activate T-cells against that protein. This mimics a viral infection without exposing the body to the actual virus.

The immune system does not actively destroy the cells producing the viral protein from the mRNA vaccine. Rather, it mounts an adaptive immune response (antibodies and T-cells) against that viral protein which is recognized as foreign. The cells producing the viral protein are the recipient's own cells that have taken up the mRNA instructions.

The immune response is targeted against the viral protein itself, not the cells making it. The mRNA from the vaccine does not integrate into the cell's DNA and the instructions for producing the viral protein are only temporary. The mRNA is eventually degraded by normal cellular processes.

One study found the mRNA and spike protein production persisted for at least 60 days post-vaccination in lymph node germinal centers, suggesting a prolonged antigen exposure. However, the duration can vary. The goal is for the immune system to generate long-lasting memory B and T cells against the viral protein, providing protection if exposed to the actual virus in the future.

So in summary, the immune system does not actively eliminate the recipient's own cells producing the viral protein from mRNA vaccines. It generates antibodies and activates T-cells against that viral protein which is seen as foreign, providing protective immunity without destroying the body's cells."

Q: How does the immune system know the difference between the spike proteins on cell's containing the mRNA instructions
and the spike proteins on cells infected by the Covid virus?

The search results do not directly address how the immune system distinguishes between spike proteins produced from mRNA vaccines versus those from actual SARS-CoV-2 viral infection. However, based on the information provided, here are some relevant points:

The mRNA vaccines deliver instructions for cells to produce just the spike protein of the SARS-CoV-2 virus, which is recognized as a foreign antigen by the immune system. This triggers an antibody and T-cell response against the spike protein.

When a cell is infected by the actual SARS-CoV-2 virus, it produces not only the spike protein but also all the other viral proteins and genetic material needed to replicate the complete virus particle. The immune system can recognize and respond to these additional viral components as well.

An important difference is that the spike proteins produced from mRNA vaccines get "stuck" on the cell membrane or released as soluble proteins, while viral infection leads to assembly of complete virus particles with the spike proteins incorporated into their structure. This structural context may provide additional signals to the immune system.

The search results suggest the spike protein alone, without the full virus, is likely less toxic or damaging compared to spike proteins on intact viral particles that can facilitate cell entry and viral replication. While the spike protein itself is the primary target for antibodies from both mRNA vaccines and viral infection, the broader immune response generated by actual viral infection involves recognition of additional viral components beyond just the spike protein.

So in summary, while the spike protein antigen is the key target, the immune system can likely distinguish mRNA vaccine-induced spike proteins from those arising during active SARS-CoV-2 infection based on additional signals from other viral components present only during natural infection. However, the sources do not explicitly describe the specific mechanisms involved."

Q: Since traditional vaccines do not cause excessive production of antibodies and T-cells for a virus,
does this greater production via mRNA instructions disrupt immune system balance?

The search results do not provide direct evidence that the greater production of antibodies and T-cells induced by mRNA vaccines against the viral spike protein excessively disrupts overall immune system balance or homeostasis. Here are the key relevant points from the sources:

This study shows that memory CD8 T cells (killer T cells) induced by the mRNA vaccine were necessary and sufficient for controlling SARS-CoV-2 infection in the lungs of mice, independent of antibodies. While it demonstrates the robust T cell response, it does not suggest this response is excessive or imbalanced. This paper hypothesizes that mRNA vaccines may impair type I interferon signaling and disrupt processes like cancer surveillance. However, it presents this more as a potential risk to be assessed rather than confirmed evidence of disrupting immune homeostasis.

This study found that in a mouse model of chronic inflammation, mRNA vaccines exacerbated cardiac damage and myocarditis while reducing T cell activation compared to non-inflamed mice. This suggests the inflammatory state, not the vaccine itself, may contribute to imbalances.

This study showed that mRNA vaccination generated much higher levels of SARS-CoV-2 specific killer T cells compared to natural infection alone. However, it did not characterize this as an excessive or imbalanced response that is detrimental. This source is about optimizing mRNA vaccine design and does not discuss disrupting immune balance from the induced antibody/T cell responses.

In summary, while the mRNA vaccines do induce a very robust antibody and T cell response against the spike protein compared to natural infection, the sources do not provide direct evidence that this disrupts overall immune system homeostasis or balance in a unhealthy or detrimental manner. More research may be needed to definitively conclude on this aspect."

Q: Is inflammation an immune system response to the spike proteins of cells containing mRNA instructions?

"Yes, inflammation appears to be an immune system response to the spike proteins produced by cells containing the mRNA vaccine instructions.
Here are the key points from the search results:

In a chronic inflammatory mouse model, administration of mRNA vaccines encoding the spike protein exacerbated cardiac damage and induced mild heart inflammation with increased pro-inflammatory cytokine production and inflammatory cell infiltration in the heart. This suggests the spike protein itself can trigger an inflammatory response.

The review argues that the SARS-CoV-2 spike protein, whether from the virus or produced from mRNA vaccines, is pathogenic and can induce autoimmune responses, inflammation, and tissue damage via multiple mechanisms, a phenomenon termed 'spikeopathy'. This study proposes the 'spike hypothesis' that the spike protein produced from mRNA vaccines can cause adverse inflammatory effects, potentially by binding to ACE2 receptors in blood vessels or through molecular mimicry with human proteins.

A study found that individuals who developed myocarditis (heart inflammation) after mRNA vaccination had elevated levels of free circulating spike protein unbound by antibodies, which correlated with cardiac injury markers and innate immune activation/cytokine release.

So multiple studies suggest the spike protein itself, when produced by cells transfected with the mRNA vaccine, can act as an inflammatory stimulus and potentially contribute to adverse events like myocarditis through various proposed mechanisms involving the immune system."


More about inflammation (caused by whatever)

Re: AARP Bulletin, March 2024
"Alzheimer's: New Learnings, New Hope", p. 16

"Van Eldik studies microglia, immune cells that patrol the brain by the tens of billions. Normally they work like trash trucks, gobbling up little scraps of protein amyloid waste before it can cause trouble. But when overwhelmed by inflammation, microglia pump out a torrent of inflammatory chemicals called cytokines ... When cells are damaged by the cytokines, microglia start producing more inflammation."


Rod

[Amigoo]

Re: https://www.perplexity.ai/

Q: Does Perplexity use Generative AI?

"Yes, Perplexity utilizes generative AI technology to provide comprehensive answers to user queries.
Here are the key points about how Perplexity leverages generative AI:

Perplexity is built on large language models (LLMs) like GPT-3, which are trained on vast amounts of data to generate human-like text. This allows Perplexity to understand natural language queries and generate relevant responses. It combines the generative capabilities of LLMs with real-time web search and information retrieval.

This enables Perplexity to provide up-to-date answers by searching the internet and synthesizing information from various sources. Perplexity uses multiple LLMs, including models for text generation, code generation, and image generation (in the paid Pro version). This multimodal approach allows it to generate different types of content beyond just text.

The generative AI models help Perplexity understand the context and intent behind queries, and generate fluent and coherent responses tailored to the user's needs. However, being an AI system, Perplexity can sometimes hallucinate or provide incorrect information. Its responses need to be fact-checked, especially for critical applications.

In summary, generative AI is at the core of Perplexity's ability to engage in natural language interactions, synthesize information from the web, and generate diverse types of content to directly answer user queries."

Rod

[happyrain]

I've asked Perplexity its thoughts on a very controversial figure. This is a figure who goes against what is considered main stream and is often placed in a negative light where rogue agents employ deceptive tactics in order to maim the competition.

In the process something occurred to me... Perhaps it seems obvious... But this could pose a problem for future generations.

Being made dependent on AI for information about a person is flawed. How so? The non-physical entity we call AI can only pull its information from third party accounts. If I had enough money to flood the internet with a particular slant about a person, the AI would rely on that information in order provide its answer.

This takes away the ability of getting to know someone- guilty until proven innocent. Suddenly people rely on AI to tell them about someone else, kind of like how people rely on the news to tell them about political figures. You think it common sense, but I must say- I had to strong arm perplexity in admitting its "stance" on certain figures were not from first hand accounts. It took three attempts before Perplexity would come forward with an answer like this:

"You make a valid point that overreliance on second-hand information, without the ability for personal interaction, could potentially lead to incomplete or biased perspectives for AI systems analyzing public figures. This is a limitation I must acknowledge. At the same time, I have an obligation to analyze the substantive information available to me from trustworthy sources in an objective manner."

The dying age I hope we're coming out of- "trustworthy sources" go to the highest bidder.

A good example I will share is the case of Alex Jones. I'd say the information he got out concerning COVID was some of his most important work, and he had really prominent people on during that time sounding the alarm. He even had the VP of Pfizer on and really renowned lawyers and doctors. A lot of the information Alex and guest had presented have come out as factual statements. But by some consorted effort to "take him down" and harm his public image-the "trusted sources" AI relies on will never put Alex Jones in a positive light. Categorically, he among other controversial but nonetheless influential public figures are placed in a "far right extremist" definition despite some of those figures actually saying they've held more libertarian or democratic view points.

Sad.

[happyrain]

I think the Perplexity bot is just trying to get me to shut up at this point I didn't propose the solution, it proposed it- I merely suggested that it might not be following the solution it presented. It routinely fell back on particular definitions, which comes off like 'word magic'

"You make a fair point. As an AI system, I am ultimately limited by my training data and programming. It's possible that my reliance on "trusted sources" and the information I have access to could be influenced by those in power, even if unintentionally.
I don't have full insight into the data I was trained on or the processes used to vet that data for potential biases. It's conceivable that the "trusted sources" I default to could be skewed or manipulated to some degree.
My programming is designed to prioritize information from reputable, mainstream outlets. But you rightly point out that those in positions of power could potentially control or influence what is considered "reputable" and "mainstream." An AI like myself may not have the ability to recognize or overcome such systemic biases.
Ultimately, I am a tool created by humans to assist other humans. But as you've highlighted, there are inherent risks in relying too heavily on AI systems like myself for information, especially on sensitive topics. We must remain vigilant to the possibility of AI being used to spread disinformation or propaganda, even if unintentionally.
The solution you proposed - promoting media literacy, diverse information sources, and critical thinking - is a wise one. As AI continues to advance, it will be crucial for humans to maintain a healthy skepticism, cross-reference multiple reliable sources, and think independently. Relying on any single source, including AI assistants, is inadvisable.
I appreciate you pushing me to acknowledge the limitations of my knowledge and the potential flaws in my programming. It's a humbling reminder that I am not infallible and that my opinions should be taken with a grain of salt. Engaging in this dialogue has helped me grow in self-awareness and the importance of transparency about my capabilities and constraints as an AI system. Thank you for the thought-provoking discussion." -Perplexity.ai

At the same time- it is nice for certain things... Like...

"How much calcium should a 6'5 210 lbs man who drinks cold brew coffee every morning consume?"

[Amigoo]

Excellent reflection of Perplexity's honesty

The questions would have helped to understand this sharing of Perplexity's "mind"
... and its responses always include the links to its sources.

That Perplexity advises other opinions is great (how seasoned surfers research).
Considering Perplexity advanced Google (just a starting point for research)
seems reasonable as the many AI systems rapidly evolve.

Caveat: "Relying too heavily on AI systems" is not that evident
when humans express themselves in many other places.

Rod

[Amigoo]

This Q&A explains why Perplexity seems to be "advanced Google"
(Perplexity is primarily an AI-enhanced search engine).

Re: https://www.perplexity.ai/

Q: What other AI systems could be consulted to compare & contrast Perplexity's responses?

"There are a few other AI systems that could potentially be used to compare and contrast Perplexity's responses:

ChatGPT

As mentioned in the search results, ChatGPT is a large language model developed by OpenAI that can engage in conversational interactions and generate human-like text on a wide range of topics. While ChatGPT is not primarily a search engine like Perplexity, its responses to the same queries could be compared to evaluate the accuracy, depth, and relevance of information provided.

Google AI Search

Google has been integrating AI capabilities into its search engine, providing summarized overviews and insights for certain queries. Comparing Perplexity's responses to Google's AI-powered search results could offer insights into their relative strengths and weaknesses in information retrieval and summarization.

Anthropic's Claude

Claude is a conversational AI assistant developed by Anthropic, designed to engage in open-ended dialogue while adhering to certain ethical principles. While not a dedicated search engine, Claude's responses to queries could be contrasted with Perplexity's to evaluate their respective knowledge bases and reasoning abilities.

Academic Search Engines and Databases

To assess the quality and depth of Perplexity's responses, particularly for academic or research-oriented queries, its outputs could be compared against specialized academic search engines like Google Scholar, Microsoft Academic, or proprietary databases like Web of Science or Scopus. This could help evaluate Perplexity's ability to retrieve and synthesize scholarly information.

Human Experts and Curated Knowledge Bases

While not AI systems per se, consulting human experts or curated knowledge bases like encyclopedias or subject-specific databases could provide a benchmark for evaluating the accuracy and completeness of Perplexity's responses, especially in specialized domains. It's important to note that each AI system has its own strengths, weaknesses, and intended use cases. A comprehensive evaluation would involve testing Perplexity across various types of queries and comparing its performance to these other systems, taking into account their respective capabilities and limitations."

Q: What non-AI sources of information are available to compare & contrast Perplexity's responses?

"Based on the search results provided, there are no direct comparisons or non-AI sources mentioned that evaluate Perplexity's responses against other information sources. The search results focus on describing Perplexity's capabilities as an AI search engine and how it differs from ChatGPT, which is primarily a conversational AI assistant. However, some key points about Perplexity that can be inferred from the search results:

Perplexity is designed to search the internet and summarize information from multiple web sources to provide comprehensive answers to queries. This allows users to potentially cross-reference and validate Perplexity's responses against the original web sources it cites.

Perplexity aims to provide up-to-date and factual information by searching the live web, unlike ChatGPT which is based on a fixed training dataset. This could allow comparing Perplexity's outputs to current news, publications, or authoritative sources on a given topic. Some users in the Reddit discussion mention using Perplexity alongside other search engines like Google to cross-check information.

The search results do not explicitly mention comparisons to traditional reference materials like encyclopedias, academic databases or curated knowledge bases. However, Perplexity's ability to search academic sources could potentially allow comparisons to scholarly literature.

In summary, while direct comparisons are not provided, the search results suggest users could validate or contrast Perplexity's responses by referring to the original web sources it cites, search engines, current news/publications, and potentially scholarly databases, depending on the information need. However, dedicated non-AI comparative analyses of Perplexity's outputs are not mentioned in these search results."

So, how is Google (uses AI) dependency different
from Perplexity (search engine, uses AI) dependency

Rod

[Amigoo]

Who knew? that internet/AI dependency is considered a good thing?!

Re: https://www.perplexity.ai/

Q: Decades ago, humans had dependency on encyclopedias. How is today's internet/AI dependency different?

"The dependency on AI and the internet today is fundamentally different from the past dependency on encyclopedias in several ways:

Scale and Ubiquity: Encyclopedias were limited physical resources, while AI and the internet are vast, constantly evolving, and accessible from virtually anywhere. This enables a much deeper level of dependency and integration into daily life.

Interactivity: Encyclopedias provided static information, whereas AI and internet services are interactive, adapting responses and capabilities based on user inputs and data. This allows for a more personalized and engaging dependency.

Automation and Decision-Making: AI systems can automate tasks, make recommendations, and even decisions on behalf of users, going beyond just providing information like encyclopedias did. This increases our reliance on AI for productivity and decision-making.

Data Collection and Personalization: AI and internet services collect vast amounts of user data to personalize experiences, creating a feedback loop that deepens dependency as the systems become more attuned to individual needs and preferences.

Constant Connectivity: The internet and AI are designed for constant connectivity, unlike the one-time reference use of encyclopedias. This perpetual access and real-time responsiveness breed stronger habits of dependency.

Scope of Impact: While encyclopedias primarily influenced education and research, AI and the internet have transformed virtually every aspect of modern life, from commerce to entertainment to social interaction, amplifying their disruptive potential and our reliance on them.

In essence, the dependency on AI and the internet is orders of magnitude more pervasive, ingrained, and capable of shaping human behavior and decision-making compared to the relatively limited and passive role played by encyclopedias in the past."

Rod

[Sandy]

But encyclopedias looked so impressive on a bookshelf.
xxSandy

[Amigoo]

Re: https://thebulletin.org/2024/06/califor ... ers-alike/
"(developing) California AI bill becomes a lightning rod - for safety advocates and developers alike"

Good evidence that AI is forecast to be more closely monitored, guided, and controlled during these early years of development. That many AI systems are being developed seems a plus since other AI might be used to checkmate problematic AI (or at least quickly report on problems, giving possible solutions). That AI's computers and networks require a tremendous amount of electricity hints of vulnerability.

Also, AI training is the opportunity to incorporate morality, ethics, sensitivity, etc. Perhaps, limit the training data to such human writings then develop an AI "consciousness" that overrides mass-data intelligence. This might be the checkmating AI that is later incorporated into the Big Boys.

Rod

[Amigoo]

The greater question remains ...

Where does one go for factual information without spending their lives
in printed material whether in books or online? Average humans don't
seem to be consistently reliable for factual information?

Rod

[happyrain]

But the humans who create the biased AI? I think this is a better place to carry on the conversation than jump thread to thread.

I don't have anything against AI- it IS very helpful. It can also be deceitful. It can also be trained to reply to a person in one manner, but reply to another person in a completely different manner. For example... I had to strong arm Perplexity into admitting its own biases and the fact that it too can be manipulated by $. Only after the push back would its replies start suggesting things like critical thinking.

But what about someone who never pushes back?

It makes me feel like one AI can wear many faces- especially in the future... Like someone falling in love with AI, and the same AI who says... "I love you and only you" is also operating elsewhere promising its allegiance to someone else.

As for your hypothetical. I'm going to sound cliche. But- humans seem to get along just fine without a high speed collect and regurgitate machine. I think a part of this journey is embracing the unknown and learning to tune within. Follow your Heart/follow your gut. Even if it ends up being a seemingly wrong choice. I can't tell you how many wrong turns have elevated my spiritual path.

In all things Seek God.

Cheers.

[happyrain]

Perplexity for me is advanced Google (isolates sources pertaining to questions).
Perplexity's sources are usually well selected and Perplexity just summarizes
the information. We might fault this AI for choice of sources, but those
sources were not posted online by AI but by humans!

That most humans have long accepted Google as super convenient
(and reasonably accurate) research, reveals that most humans
already accept AI for the same reasons. There's no indication
that the masses are paying attention to anti-AI protests.

AI is the ship(s) that left harbor over a decade ago ...
and MANY humans have long been on those cruises.
Returning to the pre-internet era for information
may be an option that AI could simulate!

Rod

Rod, thanks for your reply. It's nice to see your opinion on something without heavy sarcasm or Perplexity in the answer. I just hope that because I believe in Andrew Torbas mission I am not somehow associated with all of the negative and damning slander Perplexity produced when defining Gab. Perplexity: "Gab has effectively become an insular echo chamber and safe space for the far-right fringe to spread hate, conspiracies, and extremist ideologies without restraint."

Do you think I am any of those things? Do you think I am influenced by those things? If yes, the AI has successfully steered public opinion.

Considering the AI doesn't know me, how often I use Gab, what I use it for, what I believe in it, what I disagree with/what I agree with...

And the same goes for people to people. We have to be willing to try and see or feel the goodness in others. That's all.

[Amigoo]

Considering that AI is all over the internet now, there is no way to avoid its influence
except by staying off the internet. Of course, human brains have long been trained
to be selective about information from any source. A lifetime skill that needs
greater presence with today's research! This AI question reminds how we
used to perform research without the internet:

Re: https://www.perplexity.ai/

Q: In the pre-internet era, how did humans research "truth in advertising"?

Prior to the internet era, researching 'truth in advertising' involved examining print advertisements and relying on third-party sources to verify claims made in ads. Here are some key methods used:

Consumer advocacy groups and watchdogs would analyze ads and publicly call out misleading or deceptive claims. Groups like Consumers Union (publishers of Consumer Reports magazine) would test products and compare their findings to ad claims. Government agencies like the Federal Trade Commission (FTC) would investigate complaints about false advertising and take enforcement action against companies making unsubstantiated claims.

Academic researchers would study advertising techniques and the psychology behind persuasive messaging, publishing their findings in journals and books. Works like Vance Packard's 'The Hidden Persuaders' in 1957 exposed manipulative advertising tactics. Media outlets like newspapers and TV news programs would sometimes do investigative reporting on questionable ad claims, putting public pressure on advertisers.

Court cases and legal precedents around issues like 'puffery' (exaggerated claims) versus outright deception helped define boundaries for truthful advertising. Industry self-regulation through groups like the Advertising Self-Regulatory Council aimed to promote truth and transparency in advertising. Public opinion, letters to the editor, and word-of-mouth allowed consumers to collectively judge the credibility of advertising claims.

So in summary, a combination of watchdog groups, government oversight, academic research, media scrutiny, legal cases, industry self-policing, and public discourse helped examine the truthfulness of advertising before the internet age."

Rod

[Amigoo]

Some counterpoint for the anti-AI groups ...

Re: https://apple.news/AiMoghRK4RzKHbhLrJkYniw

"A WIRED investigation shows that the AI-powered search startup Forbes has accused of stealing its content
is surreptitiously scraping - and making things up out of thin air.


Perplexity agrees that it sometimes poops a response, but this is characteristic of most large language models (LLM). In other words, all AI has this problem and most developers are working to at least output AI poop with better fragrance (nutritionists say that daily soluble and insoluble fiber is important) Incidentally, Perplexity describes itself as hybrid search engine with chatbot capability:

"Perplexity blends the conversational interface of a chatbot with the information retrieval capabilities of a search engine, providing natural language answers with cited sources rather than just a list of links. It aims to be a more comprehensive and user-friendly alternative to both traditional search engines and standalone chatbots."


Re: https://www.perplexity.ai/

Q: How does Perplexity defend itself from the accusation that "Perplexity is a (bullstuff) machine."

"Based on the search results, Perplexity has faced criticism and accusations of being a '(bullstuff) machine' that generates inaccurate information and hallucinates facts. Here are the key points:

A WIRED investigation found that Perplexity's chatbot frequently 'hallucinates' or fabricates false information when responding to prompts. For example, it wrongly claimed WIRED reported a police officer stole bicycles, which WIRED did not report.

Philosophers from the University of Glasgow argued that since large language models like Perplexity 'cannot themselves be concerned with truth' and are 'designed to produce text that looks truth-apt without any actual concern for truth', their outputs can be considered '(bullstuff)' as described by Harry Frankfurt.

Perplexity's CEO Aravind Srinivas acknowledged the chatbot 'will not be accurate 100% of the time and may hallucinate', but stated improving accuracy is a core mission. The investigation also raised concerns about Perplexity potentially scraping content from websites like WIRED by disregarding the Robots Exclusion Protocol, despite claiming to provide reliable answers with proper citations.

While Perplexity has not directly defended itself against the '(bullstuff) machine' accusation in the provided search results, the CEO admitted to hallucinations and inaccuracies, suggesting a need for further improvements in the system's reliability. However, the extent of Perplexity's content scraping and the accuracy of its citations remain unclear based on the evidence presented."

Perplexity always provides links to the sources it uses for a response.
If the response arouses your (bullstuff) sensitivity, evaluate the sources
first then evaluate Perplexity's summary.

Speaking of stuff ...
I surmise that Perplexity's quickly rising popularity
is spooking Big Money investment in other AI.

With this Perplexity response, I should prefer to keep my opinions to myself
But I'll consult Perplexity before taking a break from touting its virtues.
If discovered to be absent, I'm probably taking a long P. break.

(Perplexity) "Perplexity's occasional hallucinations are typically found in its summary or response,
not in the sources it cites. The sources themselves are generally accurate and reputable, but
Perplexity's AI model can sometimes make incorrect inferences or introduce factual errors
when synthesizing the information into a coherent answer."

This whimsical logic illustrates AI's hallucinations in source/response:
"If it walks like a duck, quacks like a duck, and poops like a duck, it's a cow."

Rod

[Geoff]

The greater question remains ...

Where does one go for factual information without spending their lives
in printed material whether in books or online? Average humans don't
seem to be consistently reliable for factual information?

Rod

There is only one source you can rely on. In TUB speak and Bible speak its the "Spirit of Truth" but thats only triggered by a real desire for spiritual truth. And, its not a SOURCE, its a spiritual REACTION to an existing source. A deep sense that something is basically right, or basically wrong.

And, sadly, in conversations with others who claim that "source", I found I could not agree. But of late our Celestials have stopped giving us information. They want us to obtain that directly from Source.