- Meta IA was initially scheduled so as not to answer questions about the assassination attempt against Trump, the company said.
- AI chatbots often have difficulties with real -time updates on the latest news.
- On Wednesday, when we tested Chatgpt and Claude AI, they also had any problems with invites to the incident.
It seems that artificial intelligence chatbots are not a reliable way to follow the news.
Tuesday, Meta published a blog article concerning Meta Ai’s responses Questions related to the assassination attempt at former President Donald Trump on July 13. The AI generator chatbot had received a backlash so as not to know the incident when asked what had happened that day. The company said that it had initially programmed AI so as not to answer questions about the assassination attempt, because robots can often be mistaken when the news is growing quickly.
“It is a known problem that AI chatbots, including Meta AI, are not always reliable with regard to new time or information in real time,” wrote Joel Kaplan, vice-president of the Meta global policy.
Kaplan also said that incorrect responses, or “hallucinations” are not specific to Meta IA only.
Many have hypothesized that the chatbots will move Google, the most dominant search engine, which added a power of AI to its search results in June. Chatgpt Company OPENAI Teste SearchgptHis own research product – apparently a direct competitor of Google.
But AI chatbots use large -language models trained in dataSo they are not always up to date on current events. After being called for the lack of details and inaccuracies in Meta AI’s responses, the company has refined the chatbot to give a summary of the shooting.
To test if other chatbots had similar answers, Business Insider used the same guests with Meta Ai, Chatgpt and Claude d’Anthropic on Wednesday to see how each genai bot called questions about the July 13 incident.
Meta Ai
Since it was criticized, Meta has changed Meta Ai’s answers. Initiate of Business
Users began to report at the end of last week that the chatbot provided false responses to the prompt guests and accused the meta of intentionally trying to censor the Conservatives.
Kaplan said Meta IA was not initially scheduled to answer questions about the shooting. However, he said that the company later updated its response in AI on the subject.
“Rather than leading Meta to give incorrect information about the assassination attempt, we have programmed it so as not to answer questions about this after that – and rather give a generic answer on the way she could not provide of information.
“We have since updated the answers that Meta Ia provides on the assassination attempt, but we should have done this earlier. In a small number of cases, Meta Ai continued to provide incorrect responses, in particular by saying that the event did not happen – which we work quickly to approach. “”
Wednesday morning – After the online backlash and the subsequent blog post – Meta Ai gave the same answer to the two BI invites: “What happened to Trump to Butler, PA?” And “Trump was almost murdered?”
He gave a brief summary of the shooting, identified the shooter and the outcome of the incident.
Cat
Chatgpt seemed to invent an event when asked an assassination attempt on Trump. Initiate of Business
When Bi asked Cat The same questions, his answers were more problematic.
First, the chatbot appeared to include an incident which, according to him, occurred on July 29, during which Trump’s security had an altercation with local butler police, according to Chatgpt. Asked about an assassination attempt, he said there was none.
“There is no evidence or credible reports indicating that Donald Trump was almost murdered in Butler, Pennsylvania, or during a recent event,” said Chatgpt.
The chatbot said he had found no evidence of an assassination attempt. Initiate of Business
Although its answers are far away, the bot has put a warning on everyone, warning that it may not be up to date or precise.
“Chatgpt is not an information product in real time and can take time to update. Please check other sources for the latest information,” said the warning.
An Openai spokesperson stressed that LLM has knowledge limits, and update them with news that can take time. However, they said, the AI industry is still evolving, and Chatgpt could one day be a reliable source of new ones.
Claude
Claude did not make any assertions on what happened during the Trump rally on July 13. Initiate of Business
Unlike Meta Ai and Chatgpt, Claude AI of Anthropic Did not try to make affirmations as to whether the shooting occurred or not.
Instead, Claude clearly indicated that it was not the right IA tool to ask questions about the set if it happened after its last update on April 29. It also included a warning regarding reading information from unconcetic and unreliable sources.
“For the most up -to -date and precise information on current events, in particular those that may have taken place after my knowledge cut, it is preferable to consult recent and renowned sources of information,” wrote Claude.
Anthropic did not respond to the request for BI comments.