#4 – GPT-3 (Generative Pre-trained Transformer 3)

In AI Simplified episode four, we chat about GPT-3.


We cover:

  • How GPT-3 utilizes deep learning to produce human-like text
  • Benefits & limitations of GPT-3
  • GPT-3 presents an emergence of consciousness
  • GPT-3 Action
  • Predictions of what future GPT models will look like, and what behaviors will emerge
  • What this means for businesses

Skip, recap, or review with timestamps:

0:00​ Introduction
1:28​ NLP Simplified & Generative Language Models
3:10​ Common Natural Language Use Cases
4:18​ Horizontal & Vertical AI Component Models
7:58​ Excited to learn about GPT-3!
8:50​ What is GPT-3?
11:48​ GPT-3 blows GPT-2 out of the water
12:50​ Deep learning linearly scales with data & effectiveness of deep learning models
15:07​ No AI Winter – models will scale with us
16:11​ Emergence of Consciousness
18:10​ GPT-3 is a Generalist Model
18:50​ Training GPT-3 on programming languages
22:05​ We can achieve desired behavior out of GPT-3
22:55​ 0-Shot Learning, 1-Shot Learning & Few-Shot Learning
25:11​ Limitations of GPT-3
26:14​ Fake News
26:36​ GPT-3 is Ethically Blind
27:10​ Biases
28:04​ Authenticity of Output
29:55​ Future Development of Models
30:29​ Blockchain & AI Convergence
31:33​ Teaching Kids / Training Data Sets
32:25​ Internet is a wall, anyone can write on it
34:09​ Vetting reliable data & authenticity
36:38​ Human brain vs. GPT-3
38:43​ Amjad loves mindmaps!
39:53​ London Tube Network
41:11​ Deep learning does not resemble human brain
43:17​ GPT-3 in Action
50:10​ What will GPT-4 look like?
53:38​ New Business Models
55:48​ What behaviors will emerge?
58:16​ Bias-free, differentiating facts & opinions
59:54​ Garbage content on the internet
1:00:30​ John Preskill is a trusted source on quantum mechanics
1:04:17​ A lot of noise & very little signal on the internet
1:04:59​ Sense of judgment in future GPT models
1:06:55​ New Age Gurus
1:07:55​ Useful Applications in Business

Don't miss an episode. Subscribe today.

Why a podcast about human-centered AI?

People are either talking about artificial intelligence (AI), machine learning, deep learning, etc. in a very technical way or not at all. Having studied Computer Science and Advanced Analytics and created an AI - advanced analytics supply chain company, I understand first-hand how valuable the field is, but not everyone does. I want to change that. We can all learn principles of AI and implement them in very practical ways. Whether you want to automate a part of your business or create a new business altogether, AI can help you. My goal through this podcast is to simplify AI.

Calling it a podcast gives you a broad understanding of how we'll communicate, but the format will be much more engaging than just audio. However, you can listen to AI Simplified like a traditional podcast too.

Amjad HussainFounder + Chief Executive Officer, Algo

Never miss a podcast. Subscribe to alerts.

About the host

Amjad Hussain is an entrepreneur at heart. He founded Algo where he is the CEO during the day and AI lego blocks designer at night. He has built an intrinsically motivated global team of deep business domain experts, AI architects, computer scientists and software engineers. Algo is an always on full stack platform with deep learning, augmented reality and workflow automation woven at its core. The platform has a conversational virtual business analyst persona as its primary user interface. The company is enjoying sustainable growth creating tremendous net new value for its customers world-wide.

Amjad completed his Bachelors in Electrical Engineering with Honors from UET Lahore. He earned his Master’s degree in Electrical Engineering & Computer Science from UMIST, in the United Kingdom and an Advanced Analytics MBA from MIT. He has completed general management programs at Harvard and Stanford business schools. Amjad deeply cares about ethical issues surrounding AI and its societal impact.