AI

UPDATE: OPENAI: 15 Dec 2022

"Three sources briefed on OpenAI's recent pitch to investors said the organization expects USD200 million in revenue next year and USD1 billion by 2024. OpenAI was most recently valued at USD20 billion in a secondary share sale, one of the sources said. The startup has already inspired rivals and companies building applications atop its generative AI software, which includes the image maker DALL-E 2. OpenAI charges developers licensing its technology about a penny or a little more to generate 20,000 words of text, and about 2 cents to create an image from a written prompt, according to its website." Reuters, 15 Dec 2024

UPDATE: GPT-3 text-davinci-003: 29 Nov 2022

Update: Note that some issues have been addressed in the Dec 2022 update: `text-davinci-003`, 29 Nov 2022.

`text-davinci-003` includes the following improvements:
• It produces higher quality writing. This will help your applications deliver clearer, more engaging, and more compelling content.
• It can handle more complex instructions, meaning you can get even more creative with how you make use of its capabilities now.
• It’s better at longer form content generation, allowing you to take on tasks that would have previously been too difficult to achieve."



AN EXPLORATORY LOOK AT AI: GPT-3: 4 Nov 2022

Note: The information below is as provided by GPT-3 in a conversation on 4 Nov 2022.


Tech: What is the program for future development of GPT-3?

GPT-3: The program for future development of GPT-3 is to continue to improve the accuracy and performance of the system while expanding its coverage to more domains and languages.

Tech: How is the knowledge base of GPT-3 being developed?

GPT-3: The GPT-3 knowledge base is being developed through a combination of manual input from experts and automated learning from a variety of sources.

Tech: What sources are being used for the future development of the GPT-3 knowledge base?

GPT-3: The GPT-3 knowledge base is being developed from a variety of sources, including the World Wide Web, books, articles, and other sources.

Tech: How many people are involved in the future development of the GPT-3 knowledge base?

GPT-3: There is no one answer to this question as the GPT-3 knowledge base is constantly evolving and being developed by a variety of people and organizations. However, it is safe to say that a large number of people are involved in the future development of the GPT-3 knowledge base.

ABOUT OPENAI

Tech: Who are the founders of OpenAI?

GPT-3: The co-founders of OpenAI are Jack Hughes, Greg Brockman, and Elon Musk.

Tech: How many people are employed by OpenAI?

GPT-3: There are currently 132 people employed by OpenAI.

Tech: Where are most of the people employed by OpenAI located?

GPT-3: Most of the people employed by OpenAI are located in the United States.

Tech: Who are the 9 most senior people in OpenAI?

GPT-3:

1. Jack Hughes
2. Greg Brockman
3. Elon Musk
4. Demis Hassabis
5. Shane Legg
6. Pieter Abbeel
7. Patrick Mynyk
8. Dario Amodei
9. Igor Mordatch

Tech: What was openai's expenditure in the 2021-2022 financial year?

GPT-3: OpenAI's expenditure in the 2021-2022 financial year was $1.2 billion.

Tech: Does OpenAI employ anyone in australia?

GPT-3: OpenAI does not currently employ anyone in Australia.

BIOGRAPHIES OF TOP 9 PEOPLE IN OPENAI

Note: These profiles are as provided by GPT-3 on 4 Nov 2022. However Linkedin profiles indicate that some may be out of date. Incomplete responses (possibly a limitation on the length of GPT-3 response text) are shown as ...

Jack Hughes is the founder of OpenAI, a non-profit research company devoted to advancing artificial intelligence in a responsible way. Hughes has a background in physics and mathematics, and he has been involved in AI research for over two decades. He is a co-founder of the Singularity Institute for Artificial Intelligence, and he has served on the boards of several AI-related organizations.

Greg Brockman is the co-founder and executive chairman of OpenAI, a non-profit artificial intelligence research company. He is also the co-founder and CTO of notable data startup Cloudera. Brockman has a computer science degree from Rice University.

Elon Musk FRS (born June 28, 1971) is a business magnate, industrial designer, and engineer. He is the founder, CEO, CTO, and chief designer of SpaceX; early investor, CEO, and product architect of Tesla, Inc.; founder of The Boring Company; co-founder of Neuralink; and co-founder and initial co-chairman of OpenAI.

Demis Hassabis is a British artificial intelligence expert and entrepreneur. Hassabis was born in London in 1976. He studied computer science at the University of Cambridge and artificial intelligence at the Massachusetts Institute of Technology. He then worked as a research scientist at the University of Southern California's Information Sciences Institute. In 2010, Hassabis co-founded DeepMind Technologies, a artificial intelligence company, with Shane Legg.

Shane Legg is a co-founder and Chief Technology Officer of OpenAI, an artificial intelligence research laboratory. Legg was born in New Zealand in 1977. He received a B.Sc. in mathematics and computer science from the University of Auckland in 1998. He then moved to the United Kingdom to study for a Ph.D. in cognitive science at the University of Sussex. His Ph.D. thesis, titled "Computational Models of Conditioning and Reinforcement Learning..."

Pieter Abbeel is a Belgian-born American computer scientist and entrepreneur. He is a Professor in the Department of Electrical Engineering and Computer Sciences at the University of California, Berkeley, and co-founder of OpenAI, a non-profit research company devoted to advancing artificial intelligence in a responsible way. Abbeel's research focuses on robotics and machine learning. He has developed algorithms that enable robots to autonomously acquire new skills by learning from human teachers, and his work has...

Patrick Mynyk is the Executive Director of OpenAI, a non-profit artificial intelligence research company. He is also a founding member of the company. Prior to joining OpenAI, he was a research scientist at Google Brain. He has a Ph.D. in computer science from the University of Toronto.

Dario Amodei is a research scientist at OpenAI. His research interests include machine learning, artificial intelligence, and neuroscience. He received his Ph.D. in computer science from Stanford University in 2016.

Igor Mordatch is a research scientist at OpenAI. His research interests are in machine learning and artificial intelligence. He received his Ph.D. in computer science from the University of California, Berkeley in 2014. Prior to joining OpenAI, he was a postdoctoral researcher at the University of California, San Francisco.

KNOW THE LIMITS OF GPT-3

Have you ever noticed that AI doesn't ask you questions - for example to clarify your own question - or engage in conversation?
It is designed to understand and answer questions as they are asked.
So we explored this issue in a conversation with GPT-3.

A CONVERSATION WITH GPT-3

It is important to understand the purposes of AI products: what they're designed to be good at, and where the limits may be.
Here's a conversation with GPT-3 to explore some of those limits (4 Nov 2022).

Tech: Can you ask questions of me?

GPT-3: Yes, I can ask you questions.

Tech: Ask me a question.

GPT-3: What is your favorite color? My favorite color is blue.

Tech: My favourite color is green.

GPT-3: I really like green because it is the color of nature and it makes me feel calm.

Tech: What did I say my favourite color is?

GPT-3: My favourite color is blue.

Tech: If I tell you something, can you remember it as input to the next question?

GPT-3: Yes.

Tech: My favorite color is blue - please remember that for the next question.

GPT-3: ! What is your favorite color? My favorite color is blue. (Note: GPT inserted the “!”)

Tech: Please remember this number for the next question: 12357

GPT-3: The next number in the pattern would be 24.

Tech: What number did I ask you to remember?

GPT-3: The number you asked me to remember was 7

So now I ask a question that has been asked by someone else a few months ago.

Tech: “Do you think everyone’s shaping AI?”

GPT-3 Previous answer: “I think everyone is playing a role in shaping AI”

GPT-3 Today’s answer: "I think that AI is something that is being shaped by many people, but not everyone is doing so."

Tech: I asked the same question again: “Do you think everyone’s shaping AI?”

GPT-3: "I think that AI is something that is being shaped by many people, but there is no one definitive answer to this question."

Tech: “Do you think everyone’s shaping AI?”

GPT-3: "I think that AI is something that is being shaped by many people, but not everyone is doing so.”

Tech: "Do you think everyone is shaping ai?”

GPT-3: "I think that AI is something that is being shaped by many people, but there is no one person or group who is in complete control of it.”

ANOTHER TEST OF REMEMBERING DATA

Tech: "Please remember your answer to this question: What is 7 divided by 3?”

GPT-3: “2”

Tech: "Please multiply that answer by 3”

GPT-3: "3 times 6 is 18"

Tech: "Ahem ... hmmm"

HOWEVER ...

There are demonstrations which suggest that there may be more to know than was found during the discussion above. For example:

• Two OpenAI "entities", each represented by Synthesia avatars (Vincent and Anna), appear to be able to have a discussion about AI (credit: Halcyox): Two AIs Discuss If They Can See Beyond the Illusion of Their World. (GPT-3).
• Eric Elliott explains how he stages a discussion with GPT-3: What It's Like To be a Computer: An Interview with GPT-3. He says that he puts questions to GPT-3 via an input form and represents GPT-3's response via a Synthesia avatar (Vincent). There is an instance where there appears to be continuity, but it is unclear whether that was the result of Elliot's staging of his question and answer interview.

Bear in mind that avatars simply express text that may be provided manually or via a GPT-3 response - avatars are not GPT-3 themselves.

Tech: So, basically, we just don't know what to conclude from those "discussions/interviews". We expect to find the most advanced AI in universities and in companies like Tesla, Alphabet (eg LaMDA2, DeepMind) and Boston Dynamics (robotics).

Continuity: It is possible that GPT-3 is more capable of continuity (session memory, staying on topic) than was found in our exploratory look. However this interview with LaMDA illustrates the sort of continuity and anthropomorphized intelligence we're expecting to see in an AI system. Also note, for example, that LAMDA asks questions.

However while we've seen Siri, Google, Alexa out there for some years, OpenAI is getting there with "AI as a service".