• • • We Stand with Israel

• • •  

• • •  

• • • עם ישראל חי

• • • Am Yisroel Chai

• • •  

• • •  

• • • We Stand with Israel

The IA is going to take your job

Alan Hurtarte
  • NLP: Natural Language Processing
  • Interprets and comprehends the human language
  • Classifies and extract texts, besides answering questions
  • LLM: Large Language Models
  • Based on the transformer's algorithms and architectures
  • Comprehend and generate human text.
  • There is a type of model used in NLP
  • Vocodes: Systems of synthesis and analysis of voice
  • Analyze and recreates the human voice or mix it with other sounds
  • Reduce the bandwidth in communications


  • RL: Reinforcement Learning
  • With feedback, teach programs to take decisions
  • Improve abilities through iteration and solving problems
  • RLHF: Reinforcement learning from human feedback
  • Trains the model adding knowledge from experts per stage.
  • Capture better the complex human preferences
  • Diffusion models
  • Generates images from a text called prompts
  • The quality depends on the text and training data.

So the LLMs are great generative tools, but it only generates what has been learned, will never create anything new, and in most cases will be enough. But what about the other cases? When do you require to innovate and be creative? When you are solving a unique problem? Or simply looking to fulfill your client's requirements. 

How ChatGPT works

The ChatGPT model is a LLmL model. A giant one, that had millions of neural networks working together to predict the most probable outcome to your questions.

So, it’s not like ChatGPT is smart or self-aware, it is just an excellent prediction machine, between the millions of possible word combinations, calculates the most statistically correct answer. It can predict the next word of your questions, and then the next one, and the next one, and so on. That's why sometimes we get bad answers, but Chatgpt is convinced is the truth. On his database of words, it has statistical confidence that the combination of words that came as output are the right ones.

Source: https://www.scalablepath.com/data-science/chatgpt-architecture-explained

So what's next?

Probably be a couple more crazy years until the innovation flats on and we get used to it. 

Source: https://growenterprise.co.uk/2023/03/06/what-is-the-adoption-curve-of-innovation-and-how-does-it-work/

We have to look at it as a tool to be more productive, to get rid of boring and common tasks, and focus on what is important. Listen and talk to clients and users, take decisions, lead our teams, and be creative.

This could be like the internet revolution from web2.0, there will be a lot of early adopters, and the ones that could make it an integral part of the business, to the everyday tasks, and so on, will be years ahead of their competition. But we talk about a deep synergy between everyday business and IA. Automate and improve the processes.

So, it will take your job? If you have a very routine job, that can be automated, yes probably will be. But not all is bad news. When the internet become a thing, there were created a lot of new jobs like Webmaster in the early days, and now we have a lot of specialized jobs for the web. This IA disruption will create a whole new bunch of jobs and will be ready to take for the ones who prepare for it. 

If you have a job that required creativity, applying criteria, and leading teams, the IA will not take your job but definitely could help you a lot.

Alan Hurtarte
Alan Hurtarte
Lead Full Stack Developer

Light Arrow Icon
arrow icon image