Perplexity’s AI Chatbot: Why Wired Magazine Calls It a ‘BS Machine’

by

As a seasoned crypto investor with a strong background in technology and information accuracy, I find the situation surrounding Perplexity extremely concerning. The allegations of inaccurate responses, unclear mechanisms, and ethical concerns regarding data scraping are red flags that demand attention from both the tech community and investors alike.

On June 21, 2024, I, as a researcher, appeared as a guest on CNBC’s “Squawk Box” alongside Katie Drummond, Wired’s global editorial director. During our conversation, we delved into the intricacies of Perplexity, an AI search startup that had recently been the subject of an in-depth investigation by Wired magazine. The discussion brought to light several concerning aspects of Perplexity’s operations and sparked vital debates about the implications of artificial intelligence on journalism and the accuracy of information.

The Rise of Perplexity

In recent tech developments, Perplexity has emerged as a significant player, attracting substantial investments from industry leaders such as Jeff Bezos and boasting a valuation nearing a billion dollars. Despite this impressive growth, Perplexity has faced scrutiny and controversy. To clarify its role and operations, Wired conducted an in-depth investigation, examining the capabilities of Perplexity, how it functions, and the reliability of its search results.

Investigation Findings: Unclear Mechanisms and Inaccurate Responses

During her investigation for Wired, Drummond brought attention to the uncertainty surrounding how Perplexity’s technology functions. She maintains that this AI company, known for its advanced search capabilities, frequently provides inexact and possibly misleading results. Users have complained about responses that lack coherence and are heavily reliant on surface-level web data collection without further examination or comprehension.

Ethical and Legal Concerns: Scraping and Data Gathering

As a crypto investor closely following the developments surrounding Perplexity, I’ve been intrigued by the investigative findings that have come to light regarding their data gathering methods. According to reports, a substantial portion of their investigation has focused on how Perplexity obtains its data.

Biases and the Quality of AI Responses

As a researcher studying AI tools like Perplexity, I cannot stress enough the importance of acknowledging the inherent biases and the quality of information they provide. The “garbage in, garbage out” principle is a crucial concept to keep in mind. It simply means that the output of an AI model is only as accurate as the data it is given. Perplexity, for instance, heavily relies on web data which, unfortunately, can be biased or inaccurate at times. Consequently, its responses may reflect these same issues.

As an analyst, I’d put it this way: Drummond emphasized the importance of going straight to reliable journalistic sources for obtaining precise information, rather than relying on AI-generated summaries. This underscores a significant hurdle in the development of AI systems – distinguishing and prioritizing trustworthy, factual data from the vast pool of questionable sources accessible online.

The Disconnect Between Technology and Journalism

The discussion ventured into the far-reaching effects of AI on journalism. Drummond highlighted that the technology sector tends to operate independently from the moral compass guiding news and reporting. She expresses concern over how the pursuit of innovative and lucrative AI solutions can inadvertently compromise the dissemination of truthful and ethical information.

In response to accusations from Wired, Perplexity’s founder and CEO Aravind Srinivas clarified in an article for Fast Company that Perplexity doesn’t solely depend on its own web crawlers for data collection. Instead, they also utilize third-party services for web crawling and indexing. Srinivas explained that the specific web crawler identified by Wired wasn’t owned by Perplexity but by an undisclosed external provider due to a confidentiality agreement.

Srinivas acknowledged the intricacy of halting a third-party crawler from accessing Wired’s content, expressing, “It’s not simple.” He also pointed out that the Robots Exclusion Protocol, initiated in 1994, lacks legal binding and proposed that the increasing use of AI calls for a fresh form of partnership between content producers and platforms such as Perplexity.

As a researcher studying Perplexity’s answer engine, I’ve encountered criticism from Wired regarding inaccurate paraphrasing of their articles. One instance involved a false claim about a California police officer committing a crime. Srinivas explained that these results might be due to deliberately provocative prompts and assured that typical users wouldn’t face these issues. However, he acknowledged that the system isn’t error-free.

Read More

2024-06-22 09:17