Google Gemini citing sources in its response and also aiding users in dounle-checking its response.

Google Gemini | G2: Make clear how well the system can do what it can do

Google Gemini | G2: Make clear how well the system can do what it can do

Highlighting tokens with low generation probabilities (left) vs. highlighting tokens with high edit probabilities (right) for three coding tasks (Vasconcelos et al., 2023; p. 8)

Visual uncertainty highlighting | G2: Make clear how well the system can do what it can do

Alphabetize Given a list of 18 words, sort them in alphabetical order Either a missing/extra word in the resulting list, or a pair of adjacent words in the wrong order Question: Alphabetize the following words: growing prompts determining recreation evolve payable ruled patrols estimate emergency fate shrimp urges intoxicated narrator revert players pharmaceutical Answer: determining emergency evolve estimate fate growing intoxicated narrator patrols pharmaceutical payable players prompts recreation revert ruled shrimp urges Critique: Words misordered: evolve comes alphabetically after estimate

AI Critiques | G2: Make clear how well the system can do what it can do

Hugging Chat’s interface showing model selection, suggested example prompts, an input prompt box and a disclaimer “Generated content may be inaccurate or false”.

Hugging Chat | G2: Make clear how well a system can do what it can do 

Copilot in PowerPoint | 1A: Introductory blurb

Copilot in Word | G2: Make clear how well the system can do what it can do

Copilot in PowerPoint | G2: Make clear how well the system can do what it can do

Copilot in Word | 1A: Introductory blurb

Google Bard | G2: Make clear how well a system can do what it can do

Hugging Chat | G2: Make clear how well a system can do what it can do

Hugging Chat’s interface showing input examples, prompt box and a switch option for search web.

Hugging Chat | G17: Provide global controls