الأحد، 16 يوليو 2023

We tried Google’s new AI tools, and here’s what we learned



In terms of vulnerabilities, many of the new tools suffer from the same flaws that other LLM AI software tends to have. Accuracy is not guaranteed, and you should definitely double check any information you are given with an AI bot.

The quality of AI writings is also questionable. In one of the email drafts shown above, the robot started each sentence with the letter I, which every schoolboy knows is bad form. This theme is repeated in many of the emails that the bot has been asked to draft and takes time and ingenuity to fix. Likewise, it made up things that might have landed the person sending the email in hot water had it not been for editing said email first – including adding incidents not mentioned in the claim.

Another example included a draft email to a bank requesting an extension of the overdraft. We assumed that we have been with this bank for more than 10 years and have an excellent credit history. Removing this information also removed the underlying argument from the email itself.

Then there are your options. As with other tools, a successful outcome depends greatly on the quality of your router. The more information in the prompt, the more accurate Duet’s response will be. Unfortunately, that’s a lot if the tool is there to save you time, especially when you’re typing fast Talent worthy of six figures, as some employers believe. To get real utility of the tools, Google Form has to turn minimal input into something very useful.

Source link

ليست هناك تعليقات:

إرسال تعليق

AI Deepfake Ads: Tom Hanks, Gayle King Sound Warning

Tom Hanks is pretty recognizable, whether he’s holding a box of chocolates in Forrest Gump or wearing a space suit in Apollo 13. But should...