So, #ShameOnMetaAI was trending on Twitter because MetaAI was used to generate jokes on Hindu gods including Lord Ram & Lord Vishnu.
Some tweets say it declined to crack jokes on Prophet Mohammed, but others have screenshots with jokes on him.
Continue reading About offensive AI outputsYesterday, I built a custom AI chatbot to help me pack for travel, starting with making plans for an upcoming long trip to South America.
I’ve typically used spreadsheets for this, with variations based on type of travel (work, holiday), number of formal occasions, type of location (beach, mountains) among other parameters. The output is imperfect, and it does make mistakes with simple calculations, but then my spreadsheets are imperfect too: they’re all something to use as a starting point.
This is true of all AI chatbots: if you expect perfection, you will be disappointed. But if you expect a base to build on, you’ll save a lot of time.
Basic rule with AI and automation: build for small repeatable use cases and expect imperfection.
Perplexity is a fascinating AI application that I sometimes use to read the news, but it has me worried for a while now.
It takes a Wikipedia-like approach to compilation of news and information: it takes verified news sources and compiles the information in a format that is simple and easy to understand, with all the key facts. It also allows you to ask questions, answers to which it creates on the fly, based once again, on sources.
Continue reading On Perplexity and the challenge it poses for Digital MediaThe world is rapidly getting inundated with automated content: We’re seeing faceless YouTube videos grow, TV channels are deploying AI anchors. Some services that can ingest hours of someones audio, in order to generate new speech with their voice and intonation. Others allow you to get your entire body mapped to create a lifelike digital replica. The world of deepfakes is here.
The problem here is predictability and lack of personality. What often makes us interested in other humans is not the predictable part of their behavior, but what surprises us about them: what will they say, ask, or do. Some amount of predictability is important for comfort, but really what hooks us is the unpredictable.
The voices may no longer be robotic, the facial movements might now be in sync with the audio, but I’d like to believe that there are some things in a human being that are human, and inspire intrigue and trust at the same time.
So in a world of people playing it safe with AI generated content, with mass generation of how to videos made from scraping Reddit and Wikipedia, I think there’s going to be comfort in personalities, because they are both predictable and unpredictable, in predictable and unpredictable ways.
There is, of course, talk of Artificial General Intelligence that can replicate this human behavior. I’d like to believe that it can’t, for example, replace me. I suppose you’d like to believe that too.
Are AI tools a boon or a bane. On TV shows and on conference panels, I’ve spoken about how AI tools actually make our work easier, and more efficient. Whether an organisation wants to do more with the greater efficiency or make people redundant is their decision, and no fault of the AI tools. It’s important for people to learn these tools, and add these skill sets.
Two TV shows:
My show notes from the Jan 2023 discussion:
I really don’t think that journalists have anything to fear: AI tools can help them in their work, and will never entirely replace reporting. At MediaNama, we’re already using AI tools for transcription, and have started experimenting with prompts to clean copy, correct grammar, generate tweets from article, suggest article ideas, headlines, questions for interviews, speakers for conferences. We’re also creating tweets, threads, generating text to video, scripts for video. There’s so much more. The list will expand, as we become better at prompt engineering. AI will make things easier for journalists, not more difficult.
Creating and optimising videos for different platforms will be easier and faster. Processing videos will become so much more efficient.
The part that excites me most is how much translation and voice to text and text to voice will become over the next few years. That gives us the opportunity to cater to language audiences in India.
The business model challenges are of course going to increase, especially with increased advertising inventory, and fewer people required to do the same job. At the same time, this helps smaller publications punch far above their weight. How they navigate AI over the next few years is going to be critical.
A repeated some of the points I had made during the discussion, in a separate video for MediaNama: