[for-q-clinton]

Copilot is amazing for me. What does this do better than copilot for searching the web?

[dennisw]

“Copilot is amazing for me. What does this do better than copilot for searching the web?”

Ask Copilot a question then ask Perplexity the same one. Proof is in the pudding.
https://www.perplexity.ai/

[Mr. Jeeves]

Copilot keeps better track for the Deep State of which far right MAGA seditionists are asking what - and why.

None of these large public AI interfaces are "free" because their creators are feeling benevolent - they are filing away individual search patterns for future state reference.

[RoosterRedux]

Copilot is "trained" on existing data and will be biased by its training (even if that bias is unintentional). Perplexity isn't "trained" as such.

I put this issue to Perplexity and here is its response:

Rooster: I understand that Perplexity isn't trained like other ai platforms and uses machine learning to focus on current news. Is this accurate?

Perplexity.ai:

Perplexity AI is an artificial intelligence-powered conversational search engine designed to deliver accurate, real-time answers by searching the web and compiling information from multiple sources. It leverages advanced machine learning algorithms, including natural language processing (NLP), to enhance search accuracy and provide context-aware results.

Perplexity AI's main product is its AI-powered search engine, which uses NLP models like GPT-4 and custom large language models (LLMs) to understand and interpret user queries conversationally. It combines techniques such as web crawling, information retrieval, and multimodal learning to ensure its search results are based on up-to-date data.

The platform has recently introduced enhancements to its Pro Search tool, which now includes advanced mathematical and programming functionalities, allowing it to break down complex questions with step-by-step responses and rectify code errors.

Rooster: You didn't answer my question about "training." Try again.

Perplexity:

I apologize for not directly addressing your question about training. You're right to point that out, and I'll clarify: You are correct that Perplexity AI's approach differs from many other AI platforms in terms of "training."

Unlike traditional AI models that are trained on a fixed dataset and then deployed:

Perplexity AI uses a combination of large language models and real-time web search to generate responses.

It doesn't rely solely on a static knowledge base but actively retrieves current information from the internet. This approach allows it to access and incorporate up-to-date information, including recent news and events.

The system uses machine learning techniques to process and synthesize this real-time information into coherent responses.

So while it does use machine learning, its focus on current information comes from its ability to search and process web content in real-time, rather than from traditional "training" on a fixed dataset. This approach helps it stay current with recent developments and news.

[Dr. Sivana]

Copilot is amazing for me. What does this do better than copilot for searching the web?

Co-Pilot will sometimes bail out and say it is exhausted on a subject without giving an answer. A true AI is more than a large-language search engine. ChatGPT is more ambitious but also more error prone and is (by design) not current.

Perplexity isn't there either. I believe it is piggy-backing off of Google's AI, but I hope with less spying than Google or Microsoft.