This week I interviewed a Solutions Architect of a large systems integrator and discussed some aspects of building a Generative AI-driven solutions architect. The conversation was fascinating and gave some insights into the possibilities of generative AI’s impact on the workforce. If targeted models can be built to assist in a complex workload like Solutions Architect, the opportunities to expand the market for the various solutions of the companies are limitless.

During my research, I came across a blog post written by Anamaria Todor, who holds the position of Principal Solutions Architect at AWS. HellaSwag is a dataset for commonsense natural language inference (NLI) that is specifically hard for state-of-the-art models, though its questions are trivial for humans (>95% accuracy). It was introduced by Zellers et al. in their paper “HellaSwag: Can a Machine Really Finish Your Sentence?” (2019).

When it comes to Natural Language Inference (NLI), Hellaswag tests show that GPT4 models (that leverage about 1 Trillion Parameters) (Parameters are learned from a model training data) come very close to Humans (consensus is humans operate with 1 Trillion Parameters)

Source – Linkedin Blog post by Anaomorio Todor

A working paper published last week by OpenAI, the creator of GPT-4, found that 80 percent of the workforce could see at least 10 percent of their tasks performed by generative AI, based on analysis by human researchers and the company’s machine large language model (LLM).

This work is the most significant and detailed work by OpenAI, and I encourage you to read it.  The paper’s title is ” The Labor Market Impacts of Large Language Models,” it was published on arXiv on March 25, 2023.  Researchers from OpenAI, OpenResearch, and the University of Pennsylvania author it.  The study examined three categories of tasks: the direct applicability of Chat GPT, implementation through additional software investment, and integration of multiple systems using a large language model.  Although the paper is still a working draft, preliminary findings indicate a notable increase in impact as we progress from Category 1 to Category 3.  These results highlight the immense potential that large language models offer.

The most noteworthy discovery in this paper is as follows: “Our findings reveal a strong negative association between the significance of Science and Critical thinking skills and exposure to large language models (LLMs).” This suggests that occupations that heavily rely on these skills are less likely to be affected by the current influence of LLMs.  Conversely, there is a strong positive association between exposure and programming and writing skills, indicating that professions involving these skills are more susceptible to being influenced by LLMs.

This aspect is the most significant opportunity for Workforce Planners and HR leaders. 

HR leaders have a crucial responsibility in two key areas: first, in the development and implementation of AI-based workflows, and second, in reinforcing the significance of human involvement within each job.

Here are the areas where a general-purpose AGI can be beneficial in HR use cases

No one is a clear winner in this race yet, but companies are running several experiments that may be helpful for your planning.

By expanding job descriptions to clearly define and establish the collaborative roles of humans and machines, HR can harness the potential of the evolving Human-Machine world we are transitioning into.  Emphasizing the importance of human involvement alongside AI technologies will foster a symbiotic relationship, leading to increased efficiency, productivity, and innovation.