DeepSeek V3 Identifies Itself as ChatGPT

✅ DeepSeek’s New Version Identifies Itself as ChatGPT

DeepSeek recently came out with the V3, a new AI model from a Chinese lab that has been found to mistakenly identify itself as ChatGPT.

It could be trained on outputs from GPT-4, leading to issues like copying responses verbatim, resulting in inaccuracies. This practice of training models on competitors' data could compromise model quality and ethics, causing hallucinations and biases.

✅ ChatGPT Search Tool Is Vulnerable to Manipulation

A Guardian investigation found that OpenAI's ChatGPT search tool may be manipulated by hidden content on websites, leading to biased or false results.

This includes "prompt injections," where third-party instructions influence the AI's response, and malicious code returned from web searches.

Researchers warn that deceptive websites could exploit these vulnerabilities, and security improvements are needed before the tool becomes widely available.

This can be compared to the hidden keyword issues that search engines like Google faced in the past when content creators tried to manipulate search results with hidden content and misleading keywords.

🔥 Quick Hits of The Day:

OpenAI funds $1 million study on AI and morality at Duke University

OpenAI is awarding a $1 million grant to a Duke University research team to look at how AI could predict human moral judgments.

AI Needs So Much Power, It’s Making Yours Worse

With the increase in AI data across the US, it is sucking up so much power that it is impacting millions of residents across the country.

Google Search AI Overviews testing new Gemini animation

Google is pushing their AI model Gemeni as far as possible in the search and trying new animations and content styles to streamline their AI for better integration.