East Agile Works

02/03/2023

Large Language Model (LLM) developers such as OpenAI have no concern about open sourcing their algorithms.

But it is not proprietary data that is their advantage. The data sources are very public collections from Wikipedia, Reddit, Github, open web crawls, etc. And there are open source versions of this data available to just download (see The Pile).

For a little while it was not terribly hard for third parties to participate. We at East Agile spent a year training and fine tuning specialized versions of GPT models using the same structure and the same types of data as GPT-2. But we could only afford to make small versions of even GPT-2. But companies like ours were left in the dust on this training mission by GPT-3 and later larger models by the likes of Google and Nvidia and Microsoft.

The key moat is access to the computing infrastructure and piles of cash required to train the models. Developing a GPT-3 type model might easily cost $5 million in compute costs each week. Small versions that are only 70 billion parameters instead of 175 billion cost $2.5 million to train once. Maybe by the end of the process, a final model build might cost only $5 million. But that comes after countless iterations. It is hard to imagine a newbie getting it right the very first run. And GPT3.5 layers tons of expensive manual labelling on top of this process. Possibly GPT4 will cost ten times as much. Table stakes in this game are $100 million. But beware. As of just a week ago, OpenAI is now playing with $10 billion in new funding from Microsoft.

01/27/2023

AI will not put people out of work, at least not in the long run. But our roles will change. We will coordinate and manage more.

For example:

Writers will become more like editors. They will choose what to write about, how to write about it. They will validate the work. They will craft the key parts. But they will collaborate with AI when they want to deliver results very quickly. Sometimes writers will write just like they always have when they want to be poetic or very strategic. Possibly more people will produce written content, not less, in the same way that social media enables more people to publish their thoughts than ever before.

Programmers will become more like Product Managers. Programming has long been rather high level. We already often string together calls to massively complex underlying libraries. No longer are we painstakingly writing assembly language to implement things like spreadsheets. Complicated tasks are already achieved with a single line of code. With AI, the only difference will be complicated tasks are achieved with carefully worded prompts. The important high level questions will remain: what to do that will add value, generally how to do it. We will become more Product Owner paired with a team of extremely fast developers. We will still be using our brains and imagination to create results.

In other industries, such as healthcare, AI will assist doctors and nurses in diagnostics, enabling them to make more accurate and faster decisions. This will free up their time to focus on more complex cases and to provide better patient care. Similarly, in the manufacturing industry, AI will assist in automating repetitive and dangerous tasks, allowing for greater efficiency and safety for workers.

It's important to note that AI will not completely replace human workers, but rather augment and assist them in their roles. As AI becomes more advanced, it will enable us to take on more complex and higher-level tasks, allowing us to focus on more strategic and creative work. This shift will require upskilling and retraining for many workers, but it will ultimately lead to more opportunities and higher-paying jobs.

In summary, AI will change the nature of work, but it will not put people out of work. Rather, it will enhance our abilities and allow us to take on more complex and higher-level tasks, ultimately leading to more opportunities and better outcomes for businesses and workers alike.

[By Lawrence W. Sinclair, augmented by OpenAI's GPT models, Image by Dall.E2.]

01/26/2023

Many outspoken people who write frequently have criticized ChatGPT and similar language models as crutches, claiming their users are lacking in imagination or authenticity. But because of their own nature they do not recognize how helpful these tools can be for many people.

ChatGPT can provide introverted individuals with a platform to express themselves in a way that is comfortable for them, allowing them to project their ideas and thoughts more actively in written form, without the pressure of verbal communication. Additionally, ChatGPT's ability to generate text quickly and effectively can help introverted individuals overcome their fear of judgment and criticism, and to express their ideas with more confidence.

Overall, ChatGPT can be a powerful tool for both non-native English speakers and introverted individuals, to help them communicate their ideas and interests more effectively and confidently in writing.

[content augmented by AI]

01/26/2023

East Agile has for years now been working on training, fine-tuning and applying GPT class natural language models. In recent weeks these have captivated the imagination of the public with the release of ChatGPT by OpenAI.

Generative AI has been a hot topic in the tech industry for a few years now, but it seems that 2023 is finally the year that it goes mainstream. This is largely thanks to the efforts of OpenAI, a leading research organization in the field of AI.

OpenAI's mission is to ensure that artificial intelligence benefits all of humanity and is committed to making sure that their AI models are safe and can be used for the betterment of society. In recent years, the organization has made significant strides in developing cutting-edge generative models such as Dalle.E, which generates images from text, and GPT-3, and ChatGPT which has the ability to generate human-like text.

It's worth noting that OpenAI is not the only organization driving the mainstreaming of generative AI. Nvidia and Google, two of the biggest players in the tech industry, have also made significant contributions to the field.

Nvidia, a leading provider of AI hardware and software, has been instrumental in advancing the capabilities of generative models by developing powerful GPUs that are optimized for AI workloads. Additionally, the company has also released several tools and frameworks that make it easier for developers to train and deploy generative models.

Google, on the other hand, has been focused on developing and improving the performance of its own generative models. Their AI Platform provides a fully managed service for training, deploying, and managing machine learning models on a large scale. Additionally, Google AI has also released several pre-trained models like BERT, T5 and ALBERT which has been widely adopted in NLP tasks and have shown state-of-the-art performance.

Both Nvidia and Google have played a crucial role in advancing the capabilities of generative AI and making it more accessible to the general public, further driving the mainstreaming of this technology.

One of the key reasons that generative AI is going mainstream this year is the increased availability of these models to the general public. OpenAI has made ChatGPT available through a simple interface, with the promise of allowing developers and businesses to easily integrate the technology into their own applications. This has led to a surge of interest in developing innovative new products and services that leverage the power of generative AI, from software code evaluators to personalized news articles and more.

Another major factor contributing to the mainstreaming of generative AI is the increased understanding and acceptance of the technology by the general public. OpenAI has been at the forefront of educating the public about the capabilities and limitations of generative AI, helping to dispel myths and misconceptions about the technology.

Overall, it is clear that 2023 is shaping up to be a landmark year for generative AI, and we have OpenAI, Google, and Nvidia largely to thank for this. With their continued efforts in research, development and education, we can expect to see even more exciting applications of this powerful technology in the future.

East Agile plans to innovate significantly in the generative AI and machine learning fields this year.

(Most of this article was written by OpenAI's ChatGPT with guidance from Lawrence Sinclair. The AI robot image was generated by Dall.E2).

04/11/2021

In rather startling news, researchers are finding that CPUs can be four to fifteen times faster than GPUs at training Deep Learning models. This is almost the exact opposite of the current situation. Current deep learning training involves high performance matrix multiplication on GPUs. But a new approach uses hashes and search algorithms on CPUs resulting in faster performance using CPU features.

One important implication of this is that lower cost commodity cloud computing (i.e. regular AWS EC2 instances) can be used to train deep learning models faster and at lower cost. A rather large language model that might have cost $60,000 to train last year, might cost $500 to train next year when you consider speed improvements and the use of lower cost resources. We will be working with the technology at East Agile as it becomes ready for production.

This new innovation is demonstrated by Anshumali Shrivastava and Shabnam Dagahani from Rice University. Seehttps://arxiv.org/pdf/2103.10891.pdf

01/14/2020

You wouldn't be the first person to roll your eyes if told that a Silicon Valley startup that uses robots to make pizza is worth $2 billion. Nor would you be the first person to question about how that same startup, a year later, could soon double in valuation.
Read this article to see how they made it happen!

Zume Pizza is a Silicon Valley punchline, but investors are pouring millions into it.You wouldn&r...

01/07/2020

Unless you have been living under a rock for the past few years, you must have heard of the hype around AI (artificial intelligence) and ML (machine learning). These are among the most popular terms within the technology industry. Hype may come with reality, but AI/ML has no shortage of challenges. Read this article to know the surprising truth about ML and AI

Unless you have been living under a rock for the past few years, you must have heard of the hype ...

01/03/2020

User-centered design brings about many benefits in product development but it also comes with no shortage of challenges. Read this article to know top 4 challenges of user-centered design ⤵️ ⤵️

User-centered design brings about many benefits in product development (Read more: What Is User-c...

12/30/2019

No developer wants code that is difficult to work with or impossible to change. With maintainable code, the development process will be much easier in the later stages. Read this article to know why maintainable code is important and how to produce maintainable code.

For software engineers, writing code is not a problem but writing maintainable and testable code ...

12/23/2019

When your business needs a technological solution, you have to decide whether to hire a full service development company or a specialised tech team. See the advantages you will get if you choose the former.

When your business needs a technological solution, you have to decide whether to hire a full serv...