Experts believe that Artificial Intelligence (AI) will primarily perform menial tasks. Parking, grocery deliveries, and the like. For eight years straight, Elon Musk has promised full self-driving capabilities in his Tesla cars. Much progress has been made towards this goal, most other companies than Tesla. But the coughing chai while an autopilot steers us from Marina beach to Mylapore which is far away.
We may have made good — or terrible, depending on how you see it — progress elsewhere. Today, AI is writing efficient codes, creating protein and product designs, writing movies, reimagining our heavy bodies as superhero avatars. rip, make movies, play music. All of this in a fraction of the time it would take a semi-competent technician with current tools. You might consider the latest news that your plumber Perumal’s career vision is brighter than your daughter Paromita’s, who is planning a career in software engineering. So, is an AI about to take over your job?
What is AI?
It helps to get a more accurate mental model in real AI systems. Think of them like a map: from inputs to outputs. The spam detector running under Gmail is an AI that maps the content of emails into ‘spam’ or ‘ham’ categories. Dall-E, the AI animation maker, has a map from the English language in their presentation.
AI systems in Waymo vehicles (formerly Google’s self-driving car program) from current location, speed, relative position of surrounding vehicles, desired location and related information future will management work. Their non-motorized vehicles are not released in the US
Will an AI take my job?
AI historians will mark this decade as the coming age of generative AI technologies. This loosely defined term separates AI models that process images, music, and text from those that make decisions or make predictions.
Text and image generative AI models are getting better at what they do this year. The show includes a Gofundme campaign to protect artists from AI technologies that raised $150,000 this week alone. Or the World Economic Forum reports that by 2025 machines will displace 85 million jobs while they will create 97 million new jobs.
The lesson from history is that we should take those images with a grain of salt. As the advancement of AI begins in 2022, Davos is well positioned to reorganize their top 20 jobs under the reduced demand category. They have accountants, payroll clerks and bookkeeping clerks appointed on top of three. If someone loses their jobs, it’s the one who wrote the Future of Work 2020 report: accountants will do well (they have survived Excel’s best practices).
Technology rarely eliminates all tasks. In practice, these technologies simply separate tasks and reduce the work within a workplace. A farmer and his cows plowed an acre of land in a day. With a tractor, a farmer can cultivate 15 hectares in one window. Generative AI tools will introduce a parallel increase in cognitive processing. They will make some tools and skills obsolete. Such improvements in productivity and increased revenue often lead to greater productivity. This is the case in the United States. Between 2008 and 2018, among the 11 jobs considered to be at risk from AI, the number of jobs increased, on average, by 13%.
New technologies create completely new ways of working. There is no pilot before the plane. It is difficult to imagine new ways of AI development to complete. There are many ideas in that fast technology – figuring out the best way to communicate your thoughts to an AI – can become one. I was not sold. First, as noted by AI researcher Andrej Karpathy, such work is more akin to a psychologist. Second, what we need is another AI that maps from English to the quick descriptive questions that AI models want. So don’t waste your time learning the skills for a creative career.
The cost of thinking
How much does it cost to train an AI model like GPT-3? About $4.5 million, by some estimates. We cannot know for sure because OpenAI, despite its name, is very secretive about such details. Tom Goldstein, a professor of computer science at the University of Maryland, reported that it costs the company about $100,000 per day, or $3 million per month, to keep running. of ChatGPT in current demand.
At a specific level, it costs $0.0003 to publish a word out of context. If you get ChatGPT to write your MBA dissertation, in 12,000 words, it will cost less than $4 in computing costs. We also know that it costs $600,000 to train Stable Diffusion 1, a text-to-image model. If you want to use Stable Diffusion to create images with standard settings, i.e. 512 x 512 resolution, it costs $1 to buy 500 images. Design work is done by interns.
Some have proposed a three-step process to understand how humans and AI will interact in cognitive tasks in the future. From a human to a requirement, they will consult an AI system for a list of options, then take the most suitable one and refine it further for their purposes. This is a helpful plan. But note that AI does not necessarily change the way cognitive assets are created in companies.
In the future, there may be less incentive to hire and train young workers. However, I think the kids will be fine. What needs to be done is the current paradigm of information acquisition. We need to reform our education system and our traditional business as a matter of urgency. ChatGPT publishes B+ grade MBA essays on niche topics such as related contracts. High school teachers and university professors are in a bad way.
How can we advance in an AI-first world?
The answer is to be human. Characteristics such as responsibility, integrity, empathy, and patience will be valuable in the workplace. The bottom line: we often do these things in our first job. So, where do we find them – when starting as a young worker is almost impossible – is a question that I have not answered yet.
As a general rule, caring deeply for your customers and colleagues, thinking deeply about your business and its place in the world, and noticing the finer details of his knowledge base should serve others well. By writing boilerplate codes you will not get a software engineer. But start to understand latency, algorithmic complexity, and pure abstractions. Across the board, there will be less willingness to accept skilled but potentially explosive workplaces. So it’s more humble and friendly for those of us who work in white collar jobs.
The generative AI party is just getting started and the party is mint.
Write like Shakespeare, paint like Picasso
So how effective are these generative AI systems? Very good in some contexts. This photo of Marilyn Monroe and John F. Kennedy was not taken: it was created by an AI camera. It takes a little effort to spot the flaws: Monroe’s shoes and the fit of her feet; the bending of his right shoulder; the asymmetry of his collarbone. While you’re going to shun the film for its flaws, remember that we’re only a year into this technology. Also remember that the designer spends, at least, a few hours in Photoshop on something similar. Upon receiving the tip, the AI model matched this Polaroid in less than five seconds.
With the text work, the story is even wilder. In June of this year, a Google engineer who used a bad time to talk to LaMDA argued that the chatbot had been sent. The poor ship was later released. Recently, Michelle Huang, a New Yorker working at the intersection of art and technology, has adapted GPT-3, an OpenAI language model, with her own notebooks from her childhood. He said he was saved from bullying by the metaphorical question “would your eight-year-old be proud?” hearing her inner child say “I’m proud of you” through the voicemail. I asked ChatGPT, a GPT-3 optimized for conversations with people, how AI will change the future of work. Here’s the answer, with some words edited for brevity: “One thing that can be used is to help with repetitive or time-consuming tasks, allowing people to focus on mental tasks. and planning … can increase efficiency and productivity [but] should be used with human workers instead of replacing them. The use of these technologies must be carefully considered and implemented in an appropriate manner and respect the needs and interests of all employees.
Holds the presentation of Anniyan-such as personal diseases – ‘myself and ChatGP’ if it’s ChatGPT – the answer is correct, the way you expect a first year computer science graduate to come out. However, this is not the only limitation of ChatGPT. ChatGPT is good at prose, not so good at reasoning, bad at math, terrible at telling the truth. Only a combination of characteristics makes one suitable for business negotiation.
The writer is not a prophet in Colombo. He is an AI engineer.