Progress in tech products and services usually comes when a confluence of things happen at once. The invention of wearables came when sensors, the cloud, and bluetooth got combined in astounding permutations. The metaverse is being born of existing pillars: blockchain and decentralization, along with enabling technologies like NFTs, virtual reality, and artificial intelligence. Now AI is being added into a wide range of apps we used to do more manually. It’s a big deal.
A new generation of apps is being trained on vast amounts of data and learning to make intelligent decisions based on that data. The more we feed them, the more they know. Infusing AI into common tasks like writing, video-editing, healthcare, cooking, and finance is proving to be a subtle change for the end user, but ultimately will be a real game changer in how we get stuff done. Yet this game changer will not be without its serious ethical issues, either.
Thanks to the cloud and a raft of specialized AI processing chips from companies like Nvidia and Google, AI is no longer only the privilege of the big and powerful. IBM Watson, Google Assistant, Microsoft Cortana, Apple’s Siri and Amazon’s Alexa were some of the first AI infused experiences we had as consumers. Today, nearly every new app claims to use artificial intelligence
The AI-ification of Everything
According to Statista, revenue from the artificial intelligence (AI) software market worldwide is expected to reach 126 billion dollars by 2025, and grow 54% year-on-year. As per Gartner, 37% of organizations have implemented AI in some form. The percentage of enterprises employing AI grew 270% over the past four years. And according to Servion Global Solutions, by 2025, 95% of customer interactions will be powered by AI.
Today, whether it’s a dieting app like Noom which uses real-life coaches combined with AI, or a financial app like Olivia.ai which learns your about spending habits and dishes out personalized money-saving advice, the world of apps has become super-charged with AI. ELSA, short for English Language Speech Assistant, for example, uses artificial intelligence models that record users’ speech, analyze it for different benchmarks, and then come up with practice sentences to help users who want to reduce their accents.
This week I spent a lot of time looking at how AI powers a new generation of video creation tools. The problem was simple. I had hours of video to pour through and chop up into relevant snippets. Happily, AI-based solutions were plentiful. Descript starts by uploading a video and creating a transcription of its content. You might not have the skills to edit video, but imagine that you can edit the words and the video will edit itself to match your changes. Tell it take out the ums and aws, remove long pauses, make a few edits to the text so spelling is correct and you’ve got short form, good looking videos ready to publish.
Designs.ai was even more fun. I wrote a little narrative about a recent trip to Italy, and based on key words in the text it generated a movie. The program draws on 10M video clips, 170M images, 500K audio files, and 50 voices to create a fully-edited creation.
Rephrase takes your text content and matches it with a human avatar to deliver your words. The company calls itself the “MailChimp of video” because it becomes trivally easy to send out a professional quality video.
While video was my immediate AI-enhanced need, you don’t need to look far to see a cloud based world filled with little AI helpers like these. Fyle uses AI to take your receipts and immediately categorize them in the proper expense account category. Think Quickbooks but without the manual entry. Youper claims it uses AI in its chat-bot app that’s designed to help users identify, track, and process their thoughts and feelings based on cognitive behavioral therapy and other techniques. I’m a bit suspicious about this one, and the reviews are mixed. FitnessAI (iOS only) claims to create its personalized fitness programs by applying AI to your past workouts–a personal trainer for a fraction of the cost. According to the company the software’s algorithm has been trained on 5.9 million exercises. Over three years, 10 million sets, repetitions, and weights were collected from over 30,000 expert weight lifters and gym visitors.
Uncanny Valley Territory
Of course, things escalate to big-think questions and potential problems about the role of AI in our future– about privacy, making jobs obsolete or having algorithmic biases. Regulators and legistlators are already talking about labelling images that have been digitally altered, for example.
On the bleeding edge, companies like Article Forge, which comes out of research from Glimpse AI, generates articles on topics of your choosing in a few seconds.Your job is to enter your keyword, optional sub-keywords, article length, and other requirements into the Article Forge system. Not only will it write the article for you but it’ll do the SEO to make it relevant once you publish. There’s a 5 free trial and theoretically it could free you from ever having to blog on your website again. (I didn’t use it for this article, btw, but maybe we’ll give you a demo of it in coming weeks.)
Hour One creates synthetic humans based combining real human faces with your words. Record or type your narrative and then you can rent a face (in true creator-economy spirit these real people earn micropayments when their image is chosen). The program combines your words with the face of your choice. The AI matches facial expressions to the text.
The promising future of AI applications becomes even more apparent when you look at the much hyped DALL-E 2 . (Yes, that’s a play on Dali and your favorite computer generated Disney character, Wall-e.) DALL-E been in the works for years but anyone who’s seen it has seen AI’s terrifying and seductive future.
While it’s still in a very limited release to researchers and developers, you can watch videos like this one. You’ll see natural language being used to describe a scene and then that scene will be fully rendered in the artistic style of your choice. Say a “girl wearing a red dress walking up the stairs” or a “cat playing the piano” and those images magically appear. DALL-E 2 is purportedly trained on 650 million images and text captions. It’s based on the GPT3 language, one of the most advanced AI languages to date. According to the MIT Review, Meta is about to release it’s own AI language based on GPT3 speed to up the development of its platform.The language will be available to the public.
The world really is getting artificially intelligent.