AI and the Future of Work
https://futurism.com/the-byte/cnet-publishing-articles-by-ai

AI and the Future of Work

There has been a lot of excitement about Chat GPT and the implications on our society and business.? Chat GPT is an AI model classified as a Large Language Model (LLM). ?These LLMs are extremely large neural nets incorporating the concepts of attention and vector embeddings, often too large to fit on a typical computer.? The basic output of these models is to predict the most likely next word.? Their large parameter space, and the immense size of the training data, makes them difficult and expensive to train with specialized equipment.? However, once the model is trained, the use of the model can be easily implemented and leveraged to execute common writing tasks.

Criticisms

These new neural nets remain subject to the limitations all AI models are subject to, including imperfect accuracy, reliance on historical data, and generalizability.? Because AI models are statistical representations of the input data, they will always have an accuracy below 100%, and accuracies will always be lower when using the models on new concepts that were not contained in the training data (e.g., out of time data).? AI models also typically battle between identifying the specific traits of unique individual cases in the training data and finding the general rules within the data (e.g., bias vs. variance).? These are the same limitations for any AI model. ?

The latest advances in LLMs have not superseded these limitations.? As LLMs are trained on past data, the responses will only contain previously known answers to prompts.? The LLM will fail at questions that do not have a sufficient known answer.? Similarly, LLMs fail at creative prompts, as this would also entail creating new information or concepts.? This reliance on past work is a fault that creative professionals can exploit, and it is why they should not be concerned about their jobs for the time being.

Due to these limitations, the full automation of creative tasks using LLMs is not recommended.? However, the use of the technology for explanatory applications like chatbots certainly is recommended.? It is also expected that LLMs will certainly serve as useful AI assistants for many tasks and will prove an accelerant to productivity for many occupations.

Bifurcation of future work

It is also anticipated in the short run that business managers will adopt and overuse LLMs to automate many tasks formerly performed by humans. ?This will appear to dramatically increase productivity in the short run, but it is my belief that it will result in a decrease in the quality of work.? LLM automated work outputs will be less creative, as they rely on past work and is essentially re-arranging past work to fit the current query.? LLM outputs can also be of low fidelity, if the model leans more towards generalizability, which means that fine details are difficult for LLMs to reproduce.? A good example of this is the lack of detail in people’s faces in some AI generated artwork.

Due to the low cost of these LLM automated outputs, they will proliferate throughout the business world.? Every business will offer an extremely low cost and correspondingly low quality product or service that is automated by AI.? This should lead to a corresponding increase in quality of life, but only by a small amount, as the loss in fidelity and creativity will diminish the value of these products and services.

Humans will continue to be active in producing work products that compete with automated AI.? Human produced content will obviously be higher cost, but will almost certainly be of higher quality in the following ways:

·????????New and innovative content that has not been seen before

·????????New connections made between disparate concepts (so called cross-pollenization of ideas) in ways the AI cannot perform

·????????Content based on the latest ideas and concepts

This divide will proliferate to all areas of business.? Consumers will be presented with 2 main choices: 1) a low cost, AI designed, AI build product/service, or 2) a higher cost, human designed and built, high quality product/service.

While this outcome is highly likely, there is still an opportunity for creative humans to achieve much higher productivity levels given new AI technologies as AI assistants.? This strategy should also most certainly be employed as soon as possible in any business.?

Conclusion

We should expect to see this split in products and services emerge quickly.? Indeed, many customer service processes are already automated, for instance with Amazon returns.? I also believe this evolution will further enhance inequality in society in general.? Human produced goods and services will carry high enough prices to only be available to the wealthiest buyers, as in car manufacturing now, where a hand-built Bugatti commands a price far above what an average consumer can afford.? But this new advance in AI will exasperate the divide.? AI Automation will creep into more areas of our economy, and this divide between the have and the have-nots will become even more apparent.

?[No LLM was utilized in the writing of this article.]

? Copyright Lucas Finco 2024

要查看或添加评论,请登录

Lucas Finco的更多文章

  • How many EV chargers do we need?

    How many EV chargers do we need?

    I have been looking for an excuse to write an article about EV chargers, and my experiences on the NY Throughway this…

    1 条评论
  • 2020: Year of the AI Assistant

    2020: Year of the AI Assistant

    The AI revolution is here. Are you actively engaged in deploying AI in your business? If not, why not? Maybe you are…

    1 条评论
  • Electric System Peak Prediction

    Electric System Peak Prediction

    Knowing if tomorrow is going to be a peak demand day for the electric system is important to anyone who manages energy…

    2 条评论
  • Four things I want from NY REV

    Four things I want from NY REV

    The state of New York has embarked on a regulatory renovation of energy utility policies known as Reforming the Energy…

    8 条评论
  • How much energy storage does NYC need?

    How much energy storage does NYC need?

    The discussion about New York State's Clean Energy Standard and energy storage needs requires a robust analysis of…

    3 条评论
  • Diversity Benefit of DERs

    Diversity Benefit of DERs

    With the coming of Distributed Energy Resources (DERs), residential, office, and industrial buildings will be able to…

    2 条评论
  • A New Paradigm for DER Cost Effectiveness

    A New Paradigm for DER Cost Effectiveness

    The emergence of Distributed Energy Resources (DERs) and the Distributed System Platform (DSP) concept have created the…

    2 条评论
  • Why use benefit/cost testing on new energy technology incentives?

    Why use benefit/cost testing on new energy technology incentives?

    Currently, state energy regulatory agencies use incentives to promote the adoption of more energy efficient electric…

    6 条评论
  • Top 4 errors data scientists make

    Top 4 errors data scientists make

    Here are my top 4 errors that I see analysts consistently make in their application of “data science”: 1. Not having…

  • Mr. Izzo, here is how you "make it work"

    Mr. Izzo, here is how you "make it work"

    Recently, Public Service Enterprise Group Inc. (PSEG) Chairman, President and CEO Ralph Izzo responded to Pope Francis’…

    5 条评论

社区洞察

其他会员也浏览了