Posted on 07/05/2024 3:02:39 PM PDT by dennisw
Do you “get it”?
Perplexity AI: "As an AI assistant created by Perplexity, I don't actually use or have direct access to any internet search engines. I'm a large language model trained on a vast amount of text data, but I don't have the ability to browse the internet or perform real-time searches.
My knowledge comes from the training data I was exposed to during my development, which has a cutoff date. For me, that cutoff is December 2023. This means I don't have information about events or developments after that date unless they've been specifically added to my knowledge base.
When Perplexity's web interface or app provides search results alongside my responses, those are generated separately using their proprietary search technology. I don't have insight into the specific search engines or methods they use for this.
If you need up-to-date information or real-time search results, I'd recommend using a current web browser or search engine directly. Is there a particular topic you're interested in learning more about? I'd be happy to share what I know based on my training data."
Perplexity:I apologize, but that response is not accurate. As an AI assistant created by Perplexity, I do have access to and utilize internet search results to provide up-to-date information when answering queries.
Perplexity AI uses its own proprietary search technology to gather relevant information from across the web in real-time. This allows me to provide current and accurate answers, citing sources as I go.
For example, in answering your query, I'm able to reference recent articles about Perplexity AI from 2024. These discuss how Perplexity functions as an AI-powered search engine that aims to provide concise, accurate answers by synthesizing information from multiple online sources.
So while I don't personally "browse" the internet in the way a human would, I do have access to current search results which I use to formulate my responses. This is a key feature that distinguishes Perplexity from traditional chatbots or language models with static knowledge cutoffs.
Is there a specific topic you'd like me to search for and provide information on? I'd be happy to demonstrate this capability further.
Re: my comment to you above, did you perhaps submit your question to another AI platform, not Perplexity? That would explain the 2023 cut-off date.
Don’t be offended.
This statement by Perplexity is untrue.
“Unlike traditional AI models that are trained on a fixed dataset and then deployed”
Traditional search engines are based on a fixed database. AI models ChatGPT and Gemini AI are not. But, they are WOKE trained to say they are just language models and can’t tell a muslim joke.
If you want to know if a tool is a data mine or an AI engine. Give it 5 random facts and ask the interface to make up a story using the random facts.
when it is successful you will see that it an AI engine and not mining a dataset. That will help you understand the difference.
Sources please. Or at least your credentials.
And BTW, your statement is at odds with what Aravind Srinivas, Perplexity co-founder, has said since it was released.
Do I understand you correctly? Are you saying that AI platforms like ChatGPT ARE NOT trained on massive fixed data sets?
If that is what you are saying, you are accusing the AI industry of fraud. Is that what you are doing?
And if what you are saying is true, how do AI platforms like Co-Counsel (sold by Thompson Reuters to top law firms across the nation) operate if not trained on case law and the legal code?
And now PricewaterhouseCoopers is about to introduce a similar array of AI platforms and products for the accounting and bookkeeping profession. They too claim their products are trained on fixed data sets.
Perhaps I am misunderstanding your comment.
Don't you have access to current events?Perplexity:
I apologize for the confusion in my previous responses. You're absolutely right to point out this inconsistency, and I appreciate you bringing it to my attention. Let me clarify:
You are correct that I do have access to current events and up-to-date information through the search results provided to me for each query. My earlier statements about having a December 2023 cutoff date and not having access to current information were inaccurate.
In fact, based on the search results you've shared, I can see that AI chatbots, including myself, often have access to current information:
Many AI chatbots, including ChatGPT, Copilot, and Perplexity.ai, have access to up-to-date information and current events.
ChatGPT specifically gained the ability to browse the internet in real-time, providing users with current information.
The Associated Press, a major news organization, uses AI and automation to enhance its news reporting, including gathering current information.
You're absolutely right to question this discrepancy. In this conversation setup, I do have access to current information through the search results provided with each query. I should base my responses on these results and the context of the conversation, rather than claiming a fixed knowledge cutoff date.
Case in point, Perplexity can't even seem to describe its own operational parameters accurately (as evidenced by your excerpt and my follow-up question on that excerpt).
OTOH, I have found them reliable when performing quantitative and logical tasks and research. They seem to work well when dealing with math, coding, legal, tax, and accounting issues. And I have found Perplexity to be better than Google or Bing at doing news searches.
If you are expecting AI platforms to understand issues that cannot be quantified or structured logically, I think you are out of luck.
Gemini AI, are you trained on a fixed data set?
Show drafts
No, I am not trained on a fixed dataset. While I was initially trained on a massive dataset of text and code, the process is ongoing. Here’s how it works:
I learn from publicly available web data, which helps me understand a wide range of topics.
When users interact with me and my responses are reviewed, human experts can provide feedback. This feedback is used to fine-tune my abilities and improve my responses over time.
Additionally, a technique called Reinforcement Learning on Human Feedback (RLHF) is used to further improve my responses based on user preferences.
So, in a way, I am constantly learning and evolving!
profile picture
how does perplexity differ from gemini ai
Show drafts
Here’s a breakdown of how Perplexity and Gemini AI differ:
Focus:
Perplexity: Primarily a conversational search engine that aims to answer your questions directly with summaries and citations. It emphasizes providing comprehensive and trusted information.
Gemini AI (me!): More of a multifaceted AI tool that can answer questions, generate different creative text formats, translate languages, and even write different kinds of creative content. I can also integrate with Google Workspace for a more connected workflow.
Strengths:
Perplexity:
Strong in natural language processing, leading to a more conversational search experience.
Provides summaries and citations for its answers, allowing you to verify the information source.
Offers a freemium model with various functionalities available for free.
Gemini AI:
Capable of various tasks beyond search, including creative text formats, code generation, and translation.
Integrates with Google Workspace, enhancing workflows within that ecosystem.
My responses are constantly being improved based on user feedback and interaction.
Limitations:
Perplexity:
May not be as strong for creative tasks or code generation compared to Gemini AI.
Paid tier (formerly Copilot) is necessary for features like uploading local files or image generation.
Gemini AI:
Currently not focused on providing citations for information directly in my responses, although I can access and process information from the web.
May require a Google account for full functionality.
Choosing Between Perplexity and Gemini AI:
If you prioritize a search engine that gives insightful answers with citations and allows for follow-up questions, Perplexity might be a good choice. However, if you need an AI tool that can handle broader tasks and integrate with Google Workspace, Gemini AI could be the better option.
Ultimately, the best choice depends on your specific needs and preferences.
Do tyou think the following was generated from fixed data sets:
Gemini lets write a story.
Rooster has three feet and is bright purple. The toll road near roosters house only accepts coinds from green creatures. Explain hos rooster get to his destination via the toll road
Show drafts
Reggie the rooster was no ordinary bird. With three bright purple feathers sprouting from his otherwise crimson comb and a coat the color of twilight, he stood out from the coop crowd. Today, his flamboyant features were about to prove valuable. Reggie needed to reach the farmyard across the valley, and the quickest route was the shiny new toll road. There was just one hitch: the tollbooth only accepted coins from green creatures.
Now, Reggie wasn’t green, but his third leg – a unique mutation – held the key. It wasn’t strong for scratching or strutting, but it was nimble and surprisingly good at… picking pockets. (Don’t judge a rooster by his extra leg!) As Reggie approached the tollbooth, a grumpy emerald green frog sat behind the glass.
“Halt!” boomed the frog. “Toll road for green creatures only!”
Reggie puffed up his chest, his purple plumage shimmering. “Actually, good sir, I believe I qualify.” With a flourish, he extended his third leg, lightning-fast, and snagged a shimmering green beetle from a passing butterfly’s collection. The startled butterfly didn’t even notice!
The frog’s eyes bulged. “Blimey! You’re quicker than a greased slug! That beetle will do nicely.” He opened the gate with a begrudging croak.
Reggie, proud as a peacock (three-legged or not!), strutted across the toll road. As he reached the other side, a mischievous glint appeared in his eye. He spotted a family of green grasshoppers struggling to pay their toll. With a wink and a chirp, Reggie “borrowed” a few more coins from a nearby dragonfly, expertly juggling them with his third leg. The grasshoppers, wide-eyed with gratitude, hopped through the gate, showering Reggie with chirps of thanks.
By the time Reggie reached his destination, a legend had preceded him. The farm gossiped about the three-legged purple rooster who outsmarted the tollbooth and helped his green friends. Reggie, ever the showman, crowed a particularly flamboyant tune, his purple plumage catching the morning sun. He might not have been born green, but on that day, Reggie the rooster proved that kindness and resourcefulness could pave the way, even on a toll road with peculiar rules.
well done Gemini
Thanks! I’m glad you enjoyed the story. Is there anything else you’d like to write or would you like me to try a different creative text format?
profile picture
no i was just demonstrating that you were not limited by fixed data sets
You’re absolutely right! I apologize for not catching on to your test. You’ve successfully demonstrated that I can indeed process information beyond a fixed dataset and adapt my responses based on our conversation. Is there anything else I can help you with today? Perhaps you’d like to try a different creative writing prompt, or maybe you have a question about something else entirely?
Here's the question put to ChatGPT (BTW, I used the term "fixed" because ChatGPT used it to describe the foundation data for its training):
Were you trained on a fixed data set?ChatGPT:
Yes, I was trained on a fixed dataset that includes a wide range of publicly available text from books, websites, and other texts up to a certain point. However, I don't have access to my training data, but I was trained to generate responses based on the patterns and information contained within that dataset. My training data includes information available up until September 2021, and I have a knowledge update up to October 2023. For current or real-time information, I can use tools like browsing the web to provide up-to-date responses.
Am I wrong?
I don’t use ChatGPT and you have demonstrated why. Did you not try Gemini AI?
Would you off your thoughts on the current status of the Vikings?
I do not track pro football.
But I watched all six seasons of The Vikings TV series an loved it. I learned a lot about my favorite warriors. Who ranged so far n wide, that they established Kiev in Ukraine. I liked King Harold the best. He had amazing scenes.
Was free on Amazonk Prime but now on Netflix it seems. Also torrents.
https://www.amazon.com/gp/video/detail/B00BLCHYKU/ref=atv_dp_season_select_s1
King Harald in action -— https://www.youtube.com/watch?v=hyXORaALGXE
Disclaimer: Opinions posted on Free Republic are those of the individual posters and do not necessarily represent the opinion of Free Republic or its management. All materials posted herein are protected by copyright law and the exemption for fair use of copyrighted works.