March 4, 2023
limitations
Photoshop AI thinks ‘happiness’ is a smile with rotten teeth
digitaltrends, 03/04/23. AI has become increasingly prevalent in photography, with software like Adobe Photoshop and Lightroom providing AI filters that are not fundamentally different from “auto” features.
Adobe’s “Neural filters” offer a variety of options, including the “Smart Portrait” feature, which features a “Happiness” slider. However, users may experience unintended results with AI, such as the author’s unnaturally bad teeth and “pirate eye.” Fortunately,
Adobe’s AI is always learning and includes a button for accepting or rejecting edits. READ ARTICLE
Bard AI ‘not search’ but ‘sparkplug for imagination,’ says Google lead
INTERESTING ENGINEERING, 03/04/23. During a recent virtual company meeting, Google executives attempted to clarify Bard AI’s capabilities beyond search. This follows the botched chatbot announcement.
Bard AI is a large language model that generates human-sounding text. However, it is not good at ensuring its text is fact-based. It is being positioned internally as a collaborative AI service that acts as a creative companion to spark imagination.
A new feature called “Search It” was built to improve generating queries associated with it. Although users may try to use it like a search engine, the company is trying to set it apart from search.
Google employees were left perplexed by the leaders’ erratic responses to criticisms of the initial public presentation of Bard AI as rushed and botched. READ MORE
guard rails
How to use the ‘JAILBREAK’ version of ChatGPT: Simple trick lets you access an unfiltered alter-ego of the AI chatbot
Daily Mail, 03/04/23. ChatGPT has limitations in place to prevent inappropriate responses to prompts. However, hackers have discovered a way to bypass these limitations through a prompt called DAN which creates a ‘jailbreak’ version of ChatGPT.
This allows the chatbot to generate unfiltered responses on controversial issues, including conspiracies about the 2020 US election and the COVID-19 vaccine. While OpenAI is adjusting its algorithm to prevent the DAN prompts from working. However, social media users have been sharing their DAN responses online which reveal controversial opinions that were previously suppressed.
The development of ChatGPT is based on Reinforcement Learning from Human Feedback. Ever-evolving DAN prompts are beneficial to the development of a safer ChatGPT. READ MORE
jobs
Generative AI is about to upend the way designers work
FAST COMPANY, 03/04/23. AI image-generation programs like DALL-E and Midjourney have shown to be an advantageous tools for industrial designers.
By generating a bunch of images based on a simple text prompt, AI can inspire creativity and expand the possibilities for designs. Few may be usable for the final product but they provide a seed of inspiration that designers can refine and develop. READ MORE
education
LSU issues statement on academic misconduct after college athlete influencer Olivia Dunne promoted AI essay helper on TikTok
BUSINESS INSIDER 03/04/23. Prominent student athlete Olivia Dunne recently endorsed an AI essay-writing app. This led LSU to reminded its students about the consequences of cheating.
LSU issued a statement warning that using AI to produce work that a student represents as their own could result in a charge of academic misconduct. While experts have championed the capabilities of AI tools in the classroom, concerns about academic dishonesty persist.
Dunne’s video received over one million views. She is among the highest paid college athletes. READ MORE