Companies scramble to capitalize on generative models like ChatGPT, what is the real impact on jobs and the economy?
In recent months, an artificial intelligence gold rush has been underway, spurred on by the promise of generative AI models such as ChatGPT.
This technology, released by OpenAI in November 2021, has captured the attention of app developers, venture-backed startups, and even some of the largest corporations in the world.
As businesses look for ways to capitalize on this new technology, it is important to consider the impact it will have on workers and the economy as a whole.
While there is no denying that generative AI models like ChatGPT have the potential to automate tasks that were once thought to be solely in the realm of human creativity and reasoning, such as writing, creating graphics, and analysing data, their limitations are also clear. One of the chief concerns is their propensity for making stuff up. This raises questions about their reliability and accuracy, and how much they can truly be trusted in certain contexts.
The potential impact of generative AI models on the workforce and overall productivity is also uncertain. While automation has the potential to increase efficiency and reduce costs, it can also lead to job loss and economic instability.
This is particularly true in industries where the use of AI could displace human workers, such as in journalism or content creation.
Despite the amazing advances in AI and other digital tools over the last decade, their record in improving prosperity and spurring widespread economic growth is discouraging.
While some investors and entrepreneurs have become very rich, the benefits have not been shared equally. In fact, many people have been left behind, with some even being automated out of their jobs. It is important to consider the potential consequences of new technologies like ChatGPT, and to work to ensure that they are used in ways that benefit society as a whole.
Which trends will steer society in this context? Let’s explore some probable events that will happen to modern societies worldwide around ChatGPT:
- More companies invest in generative AI models like ChatGPT to automate tasks previously done by humans, leading to job displacement in certain industries.
- ChatGPT and other AI models become more advanced and reliable, increasing the potential for automation across various industries.
- Concerns about the ethical implications of AI-powered automation grow, leading to calls for more regulation and oversight.
- Tech companies like OpenAI, which developed ChatGPT, become increasingly influential in the global economy as their products and services gain popularity.
- Startups emerge with new products and services based on generative AI models, disrupting traditional industries and creating new job opportunities.
- Governments and educational institutions begin to prioritize AI education and training programs to help workers adapt to the changing job landscape.
- Some workers resist automation and push back against companies that are implementing AI solutions, potentially leading to labour disputes.
- Companies that effectively implement AI-powered automation see increased profits and productivity, leading to more pressure on other businesses to follow suit.
The use of AI-powered automation becomes a key issue in political debates and elections, with candidates proposing various solutions to address the potential impacts on jobs and the economy.
- New industries and business models emerge that are specifically designed to harness the power of AI, such as data analytics and predictive modelling.
Impact on education
Despite the uncertainties surrounding the impact of generative AI models on education and the economy as a whole, one thing is clear: the amazing advances in AI and other digital tools over the last decade have not yet translated into widespread economic growth or improved prosperity. While a few investors and entrepreneurs have become very rich, most people have not benefited and some have even been automated out of their jobs. As the gold rush for AI continues, it remains to be seen whether the technology will ultimately be a force for good or a source of further inequality and disruption.
The true revolution in artificial intelligence is yet to come: the (intelligent) automation levels
From school children to recipe developers to search engines, AI has become increasingly accessible and ubiquitous, and I myself am quite enthusiastic about the potential for generative AI tools to benefit humanity. However, it’s important to note that such tools still require human intervention and assistance to truly succeed in any endeavour. While ChatGPT may be able to assist in writing an essay, it would be unwise to rely on it to automatically respond to emails, as it is primarily a suggestive tool by design.
The development of generative AI tools is the result of decades of work in the industry, and the increasing popularity of apps such as Stable Diffusion and ChatGPT have made these tools more accessible than ever. This accessibility is a significant advantage of ChatGPT, as anyone can experiment and have fun with it.
For founders and entrepreneurs interested in using AI to create groundbreaking products and services, it is crucial to carefully consider the goal of your invention. Is it to complete a task without humans or to assist humans with a task? Take vehicle autonomy, for example.
We categorize vehicles today into several levels of autonomous driving capabilities. Level 1 offers some automated assistance, such as cruise control, while Level 2 includes partial automation, such as automatic braking and parking assistance (steering). Both GM’s Cadillac Super Cruise and Tesla’s Autopilot systems function at Level 2.
Level 3 involves environmental-detection capabilities, such as passing a slow vehicle, and the 2019 Audi A8L was the first Level 3 car in production (authorized in Europe only). Vehicles at this level require an alert, trained human driver to be able to take over if conditions for optimal self-driving safety cannot be met.
Most generative AI techniques operate like Level 1-3 autonomous driving capabilities, assisting the human driver but still requiring a human because they are not accurate or sufficient enough to operate without one. They are often called “Advanced Driver Assistance Systems” or ADAS.
The next two groups of autonomous capabilities require an evolutionary shift in the use of AI technologies because the results must be error-free or almost error-free. After all, a red traffic light is not a “suggestion” to stop. And we do not want a vehicle “deciding” to race another car or something similarly hazardous.
Level 4 vehicles, for example, operate entirely in self-driving mode without a human operator behind the steering wheel. Until legislation and infrastructure are implemented, these vehicles are typically allowed only in specific areas, and Cruise robotaxis operate at this level, serving San Francisco, Austin, and Phoenix. Level 5 autonomy operates without a precomputed high-definition map, meaning the vehicle is less dependent on prior human inputs.
I believe it is helpful to use similar levels of autonomous driving capabilities – Level 1 to Level 5 – in evaluating the capabilities of AI technologies. As I mentioned earlier, most generative AI techniques require a human operator, making them a Level 1.
Yann LeCun, the inventor of convolutional neural networks and Chief AI Scientist at Meta, acknowledges that generative AI models like Stable Diffusion and ChatGPT are impressive, but limited. I concur, as applying AI technology without considering its appropriate use can lead to problems.
While AI is not a universal remedy, it can be useful for certain situations, but it is essential to use the correct type of AI for the given scenario. For example, generative AI technologies may not be the best option for products that require precision and adherence to regulations, while dealing with unpredictable events and inputs.
To develop products such as Level 4 and 5 self-driving cars, more robust AI is necessary to handle uncertainty and scale effectively.
The leap from human-assisted products to those that can operate autonomously is significant.
