What is OpenAI, ChatGPT and DeepMind?

 (originally posted on Medium on Dec 11, 2022)


The euphoria around solutions such as OpenAI’s ChatGPT as well as Google DeepMind’s AlphaCode is unlike anything I’ve seen for quite a while. Here are some of the statements that are being made:

“OpenAI’s ChatGPT is the Google Search killer.”

“OpenAI’s ChatGPT and Google’s DeepMind AlphaCode will completely change how programming is done in the future.”

What is OpenAI?

Open AI is a research/development company focused on Artificial Intelligence. “OpenAI’s mission is to ensure that artificial general intelligence (AGI) — by which we mean highly autonomous systems that outperform humans at most economically valuable work — benefits all of humanity.”

There are many heavy hitters behind OpenAI. “OpenAI LP is governed by the board of the OpenAI nonprofit, comprised of OpenAI LP employees Greg Brockman (Chairman & President), Ilya Sutskever (Chief Scientist), and Sam Altman (CEO), and non-employees Adam D’Angelo, Reid Hoffman, Will Hurd, Tasha McCauley, Helen Toner, and Shivon Zilis. Our investors include Microsoft, Reid Hoffman’s charitable foundation, and Khosla Ventures.” Big guns for sure. Elon Musk and Sam Altman founded OpenAI back in 2015.

OpenAI has released the following artificial intelligence capabilities:

  • ChatGPT — a trained a model which interacts in a conversational way. The dialogue format makes it possible for ChatGPT to answer followup questions, admit its mistakes, challenge incorrect premises, and reject inappropriate requests. (released Nov 30, 2022) A list of documents and APIs can be found here… https://github.com/humanloop/awesome-chatgpt
  • DALL.E 2 — AI system that can create realistic images and art from a description in natural language. 3M+ users generating 4M images per day.
  • Whisper — open-sourced a neural net that approaches human level robustness and accuracy on English speech recognition

What is Google DeepMind?

Google DeepMind is Google’s artificial intelligence subdivision based in London.

Google and DeepMind have released the following artificial intelligence capabilities:

  • AlphaCode — uses AI to write computer code. Recently ranked in the top 54% of human programmers in a coding competition, and achieved “approximately human-level performance”. AlphaCode was designed specifically for competitive programming.
  • AlphaFold — can accurately predict 3D models of protein structures and has the potential to accelerate research in every field of biology
  • WaveNet — creates more natural-sounding speech for products used by millions of people around the world
  • Imagen (specifically from Google Research group) — a text-to-image diffusion model with an unprecedented degree of photorealism and a deep level of language understanding
  • LaMDA — Google’s AI chatbot — Language Model for Dialogue Applications. It mimics speech by ingesting trillions of words from the internet.

What is Github CoPilot?

Github CoPilot uses the OpenAI Codex to “suggest code and entire functions in real-time, right from your editor.” https://github.com/features/copilot

CoPilot provides coding suggestions in numerous languages. Github ran the following test and the results were compelling. (details in the link below)

Github’s Copilot test

General thoughts…

  1. If you are in the technology space, carve out time to keep abreast of the advancements in AI. Get familiar with OpenAI’s and DeepMind’s APIs. Write some code!
  2. Consensus is that we are far off from AI being able to create reliable complex code, but using solutions like Github’s Copilot are proving that the developer experience can be improved by leveraging a simple code suggestion capability. Have some of your developers check it out and determine whether it helps improve their coding experience.
  3. If solutions like OpenAI’s GPTChat were to unseat Google Search, everything from marketing to advertising to commerce would change. Let’s face it — Google owns us. If we want to purchase something (Search), learn something (Search+Youtube), browse (Chrome), get somewhere (Maps/Waze), read the news (News) or collaborate (Drive, Meet, Workspace), then many of us GLOBALLY use Google.
  4. The creation of art work and images will heavily leverage AI. Design, advertising and marketing will expand to leverage AI-generated material. The ability to create truly unique images with incredible image quality is now a one-click capability.
  5. There are numerous concerns being raised about AI. Is AI making decisions or recommendations based on racial, gender or other types of biases? Are the AI models built on data that is skewed? Are capabilities like facial recognition impacting people’s privacy and security? How do we detect and address deepfake videos that are designed to mislead people?
  6. Other AI solutions include image generators like Lensa and Stable Diffusion. They are similar to OpenAI’s Dalle2.

Comments

Post a Comment

Popular posts from this blog

Enterprise Architecture Inside Salesforce

Image
  Enterprise  Architecture Inside Salesforce Enterprise Architecture — How We Do It Internally The two words, “Enterprise Architecture”, elicit various reactions — everything from “ we tried that and it doesn’t work ” to “ we need that! ” to “ we have an effective EA practice ”. Unfortunately it is possible (even easy) to do Enterprise Architecture  incorrectly , with an approach that is lethargic, technology-first and theoretical — the classic ivory tower. (In Salesforce, the internal IT group reporting into our CIO is called “Business Technology” so you’ll see me use BT to refer to Salesforce IT) I lead the Salesforce Enterprise Architecture team within Business Technology and I’m often asked the following questions: how does Salesforce do Enterprise Architecture? why do you need Enterprise Architecture? how do you shift the conversation to business capabilities? Illuminate — The “How” of Salesforce Enterprise Architecture Let me start with the “HOW” first. Within Sales...
Read more

Illuminate - Business Model Exploration

Image
  Illuminate — Business Model Exploration One of our world-class Salesforce Enterprise Architects, Jonathan Laudicina ( https://www.linkedin.com/in/jonathan-laudicina/  ) has done amazing work on Business Model Exploration. I want to highlight some of his work and I think you’ll find it useful. As a reminder, Illuminate is the Enterprise Architecture Methodology that we use inside Salesforce as well as with our customers. The overall Illuminate methodology is depicted below. You will notice that Business Model Exploration is the first step in Illuminate. Business Model Exploration is the methodology we use with customers to articulate: External Forces pushing on the business Internal Forces pushing on the business Customer specific Business Model Canvas At the heart of Business Model Exploration is the Business Model Canvas. It is the place in which you capture key factors about a company’s specific value proposition, partners, customer segments, cost structure and revenue str...
Read more

5 Key Components of a Successful Enterprise Architecture Function

  5 Key Components of a Successful Enterprise Architecture Function 5 Key Components of a Successful Enterprise Architecture Function Creating and managing a successful Enterprise Architecture function requires a variety of different hard and soft skills. In addition, each company is different and the Enterprise Architecture function needs to calibrate and align itself to the specific company. However, there are five common features of a successful Enterprise Architecture function that are applicable to all companies. 1. Governance Enterprise Architecture (EA) requires governance, however not in the form of complex documents, forms or processes.  Definitely avoid EA being seen as the policy-enforcement team — people will avoid EA if the perception is that EA is a blocker.  The best governance starts with a simple, recurring dialogue across multiple functions facilitated by EA. I am always amazed at how valuable it is just to get cross-functional teams to  talk . Pick...
Read more