How AI Will Change Work
- By Paul Mah
- November 09, 2022
The buzz around AI continues to grow rapidly, as new research and AI models are released by the week. Just last week saw Google finally offer public access to its text-to-image AI system, while Meta has also recently unveiled a new AI-powered language translation service offering a jaw-dropping 200 languages.
Elsewhere, a recent report by analyst firm Forrester says off-the-shelf and custom AI software will grow 50 percent faster than the overall software market over the next two years. With the breakneck pace of development in AI showing no sign of abating any time soon, how can we expect AI to impact us professionally?
Generative AI is changing our world
While some have pointed out that the current approach to AI is limited and will not get us to artificial general intelligence (AGI), others such as Sonya, a partner at Sequoia, have noted that the generative AI hype is real and will have an impact on us today.
In an interview with Protocol, Huang thinks this is due to a combination of market maturity as AI models improve to the point of human-level performance. Crucially, access to many of these models is now broadly available, compared to the closed betas and limited access of the past few years. This has led to a wave of new business-centric capabilities being designed and built.
Of course, there are also unintended consequences. Take US-based illustrator Hollie Mengert, who woke up one morning to find that someone had taken 32 of her illustrations and used them to fine-tune Stable Diffusion to – you guessed it – generate new illustrations in her unique style. This was subsequently released under an open license for anyone to use.
According to Andy Baio of Waxy who reported on the issue, it turns out that this was the work of a Nigerian mechanical engineering student in Canada who decided to lend a hand to a fellow Reddit user. By making use of the free-to-access Stable Diffusion, the training process took all of 2.5 hours and cost less than USD2 on a rented GPU on Vast.ai.
According to Baio, there are now over 700 models in the Concepts Library on HuggingFace so far, with models based on classic Disney animated films, modern Disney animated films, Cyberpunk, K-pop singers, and many others.
The tipping point of AI
There is no question that generative AI models are disrupting the field of art, including winning art competitions. And soon enough, it might well impact many other fields, such as video creation, coding, image manipulation, speech generation, and even copywriting, to name a few.
Of course, it is worth noting that smartphones and modern web browsers on laptops and desktops didn’t just appear overnight. It took decades for the hardware to be developed and software refined, and many years more before pervasive access changed the equation. The question is: are we at the tipping point for AI yet?
Writing in a Bloomberg opinion column a couple of weeks back, columnist Tyler Cowen and professor of economics at George Mason University thinks this tipping point might soon be here. He wrote: “AI is about to revolutionize our entire information architecture. You will have to learn how to use the internet all over again.”
At the heart of Cowen’s argument is how a new generation of AI-powered systems will soon become the de facto information aggregator of the future. Just like how Facebook became our go-to for news, a new generation of AI-powered systems could soon determine what we see and read.
In Cowen’s view, getting ahead in this brave new world might require us to learn new AI-relevant skills to replace that of “Twitter”, “SEO” or “social media”.
Time to learn new skills
For now, I don’t think AI will completely replace humans yet. However, a growing list of AI-powered tools will compel professionals in various fields to use them to compete effectively. Just like how graphic designers took to using computers, programmers, designers, and writers will soon need to start using AI tools to stay relevant.
If you have yet to try your hand at writing the prompts used to generate AI art, let me assure you that it is harder than it looks. Part programming and part writing, it takes a fair amount of effort to get text-to-image systems to generate the artwork that you want. Which is why you will need to roll up your sleeves and practice.
If you are not worried yet, Huang of Sequoia shared a list of examples in a viral tweet, offering a glimpse of a rapidly growing field that ranges from code generation, technical documentation, video editing, text to SQL, and more. Check it out and let me know your thoughts.
Paul Mah is the editor of DSAITrends. A former system administrator, programmer, and IT lecturer, he enjoys writing both code and prose. You can reach him at [email protected].
Image credit: iStockphoto/choochart choochaikupt
Paul Mah
Paul Mah is the editor of DSAITrends, where he report on the latest developments in data science and AI. A former system administrator, programmer, and IT lecturer, he enjoys writing both code and prose.