GPT Store To Launch Next Year, ChatGPT Updates Coming Soon via @sejournal, @kristileilani

OpenAI shares its plans for the GPT Store, enhancements to GPT Builder tools, privacy improvements, and updates coming to ChatGPT.

• OpenAI has scheduled the launch of the GPT Store for early next year, aligning with its ongoing commitment to developing advanced AI technologies.
• The GPT Builder tools have received substantial updates, including a more intuitive configuration interface and improved file handling capabilities.
• Anticipation builds for upcoming updates to ChatGPT, highlighting OpenAI’s responsiveness to community feedback and dedication to AI innovation.

Researchers Extend GPT-4 With New Prompting Method via @sejournal, @martinibuster

Microsoft published a research study that demonstrates how advanced prompting techniques can cause a generalist AI like GPT-4 to perform as well as or better than a specialist AI that is trained for a specific topic. The researchers discovered that they could make GPT-4 outperform Google’s specially trained Med-PaLM 2 model that was explicitly trained in that topic.

Advanced Prompting Techniques

The results of this research confirms insights that advanced users of generative AI have discovered and are using to generate astonishing images or text output.

Advanced prompting is generally known as prompt engineering. While some may scoff that prompting can be so profound as to warrant the name engineering, the fact is that advanced prompting techniques are based on sound principles and the results of this research study underlines this fact.

For example, a technique used by the researchers, Chain of Thought (CoT) reasoning is one that many advanced generative AI users have discovered and used productively.

Chain of Thought prompting is a method outlined by Google around May 2022 that enables AI to divide a task into steps based on reasoning.

I wrote about Google’s research paper on Chain of Thought Reasoning which allowed an AI to break a task down into steps, giving it the ability to solve any kind of word problems (including math) and to achieve commonsense reasoning.

These principals eventually worked their way into how generative AI users elicited high quality output, whether it was creating images or text output.

Peter Hatherley (Facebook profile), founder of Authored Intelligence web app suites, praised the utility of chain of thought prompting:

“Chain of thought prompting takes your seed ideas and turns them into something extraordinary.”

Peter also noted that he incorporates CoT into his custom GPTs in order to supercharge them.

Chain of Thought (CoT) prompting evolved from the discovery that asking a generative AI for something is not enough because the output will consistently be less than ideal.

What CoT prompting does is to outline the steps the generative AI needs to take in order to get to the desired output.

The breakthrough of the research is that using CoT reasoning plus two other techniques allowed them to achieve stunning levels of quality beyond what was known to be possible.

This technique is called Medprompt.

Medprompt Proves Value Of Advanced Prompting Techniques

The researchers tested their technique against four different foundation models:

1. Flan-PaLM 540B
2. Med-PaLM 2
3. GPT-4
4. GPT-4 MedPrompt

They used benchmark datasets created for testing medical knowledge. Some of these tests were for reasoning, some were questions from medical board exams.

Four Medical Benchmarking Datasets

1. MedQA (PDF)
   Multiple choice question answering dataset
2. PubMedQA (PDF)
   Yes/No/Maybe QA Dataset
3. MedMCQA (PDF)
   Multi-Subject Multi-Choice Dataset
4. MMLU (Massive Multitask Language Understanding) (PDF)
   This dataset consists of 57 tasks across multiple domains contained within the topics of Humanities, Social Science, and STEM (science, technology, engineering and math).
   The researchers only used the medical related tasks such as clinical knowledge, medical genetics, anatomy, professional medicine, college biology and college medicine.

GPT-4 using Medprompt absolutely bested all the competitors it was tested against across all four medical related datasets.

Table Shows How Medprompt Outscored Other Foundation Models

Why Medprompt is Important

The researchers discovered that using CoT reasoning, together with other prompting strategies, could make a general foundation model like GPT-4 outperform specialist models that were trained in just one domain (area of knowledge).

What makes this research especially relevant for everyone who uses generative AI is that the MedPrompt technique can be used to elicit high quality output in any knowledge area of expertise, not just the medical domain.

The implications of this breakthrough is that it may not be necessary to expend vast amounts of resources training a specialist large language model to be an expert in a specific area.

One only needs to apply the principles of Medprompt in order to obtain outstanding generative AI output.

Three Prompting Strategies

The researchers described three prompting strategies:

1. Dynamic few-shot selection
2. Self-generated chain of thought
3. Choice shuffle ensembling

Dynamic Few-Shot Selection

Dynamic few-shot selection enables the AI model to select relevant examples during training.

Few-shot learning is a way for the foundational model to learn and adapt to specific tasks with just a few examples.

In this method, models learn from a relatively small set of examples (as opposed to billions of examples), with the focus that the examples are representative of a wide range of questions relevant to the knowledge domain.

Traditionally, experts manually create these examples, but it’s challenging to ensure they cover all possibilities. An alternative, called dynamic few-shot learning, uses examples that are similar to the tasks the model needs to solve, examples that are chosen from a larger training dataset.

In the Medprompt technique, the researchers selected training examples that are semantically similar to a given test case. This dynamic approach is more efficient than traditional methods, as it leverages existing training data without requiring extensive updates to the model.

Self-Generated Chain of Thought

The Self-Generated Chain of Thought technique uses natural language statements to guide the AI model with a series of reasoning steps, automating the creation of chain-of-thought examples, which frees it from relying on human experts.

The research paper explains:

“Chain-of-thought (CoT) uses natural language statements, such as “Let’s think step by step,” to explicitly encourage the model to generate a series of intermediate reasoning steps.

The approach has been found to significantly improve the ability of foundation models to perform complex reasoning.

Most approaches to chain-of-thought center on the use of experts to manually compose few-shot examples with chains of thought for prompting. Rather than rely on human experts, we pursued a mechanism to automate the creation of chain-of-thought examples.

We found that we could simply ask GPT-4 to generate chain-of-thought for the training examples using the following prompt:

Self-generated Chain-of-thought Template

## Question: {{question}}
{{answer_choices}}
## Answer
model generated chain of thought explanation
Therefore, the answer is [final model answer (e.g. A,B,C,D)]"

The researchers realized that this method could yield wrong results (known as hallucinated results). They solved this problem by asking GPT-4 to perform an additional verification step.

This is how the researchers did it:

“A key challenge with this approach is that self-generated CoT rationales have an implicit risk of including hallucinated or incorrect reasoning chains.

We mitigate this concern by having GPT-4 generate both a rationale and an estimation of the most likely answer to follow from that reasoning chain.

If this answer does not match the ground truth label, we discard the sample entirely, under the assumption that we cannot trust the reasoning.

While hallucinated or incorrect reasoning can still yield the correct final answer (i.e. false positives), we found that this simple label-verification step acts as an effective filter for false negatives.”

Choice Shuffling Ensemble

A problem with multiple choice question answering is that foundation models (GPT-4 is a foundational model) can exhibit position bias.

Traditionally, position bias is a tendency that humans have for selecting the top choices in a list of choices.

For example, research has discovered that if users are presented with a list of search results, most people tend to select from the top results, even if the results are wrong. Surprisingly, foundation models exhibit the same behavior.

The researchers created a technique to combat position bias when the foundation model is faced with answering a multiple choice question.

This approach increases the diversity of responses by defeating what’s called “greedy decoding,” which is the behavior of foundation models like GPT-4 of choosing the most likely word or phrase in a series of words or phrases.

In greedy decoding, at each step of generating a sequence of words (or in the context of an image, pixels), the model chooses the likeliest word/phrase/pixel (aka token) based on its current context.

The model makes a choice at each step without consideration of the impact on the overall sequence.

Choice Shuffling Ensemble solves two problems:

1. Position bias
2. Greedy decoding

This how it’s explained:

“To reduce this bias, we propose shuffling the choices and then checking consistency of the answers for the different sort orders of the multiple choice.

As a result, we perform choice shuffle and self-consistency prompting. Self-consistency replaces the naive single-path or greedy decoding with a diverse set of reasoning paths when prompted multiple times at some temperature> 0, a setting that introduces a degree of randomness in generations.

With choice shuffling, we shuffle the relative order of the answer choices before generating each reasoning path. We then select the most consistent answer, i.e., the one that is least sensitive to choice shuffling.

Choice shuffling has an additional benefit of increasing the diversity of each reasoning path beyond temperature sampling, thereby also improving the quality of the final ensemble.

We also apply this technique in generating intermediate CoT steps for training examples. For each example, we shuffle the choices some number of times and generate a CoT for each variant. We only keep the examples with the correct answer.”

So, by shuffling choices and judging the consistency of answers, this method not only reduces bias but also contributes to state-of-the-art performance in benchmark datasets, outperforming sophisticated specially trained models like Med-PaLM 2.

Cross-Domain Success Through Prompt Engineering

Lastly, what makes this research paper incredible is that the wins are applicable not just to the medical domain, the technique can be used in any kind of knowledge context.

The researchers write:

“We note that, while Medprompt achieves record performance on medical benchmark datasets, the algorithm is general purpose and is not restricted to the medical domain or to multiple choice question answering.

We believe the general paradigm of combining intelligent few-shot exemplar selection, self-generated chain of thought reasoning steps, and majority vote ensembling can be broadly applied to other problem domains, including less constrained problem solving tasks.”

This is an important achievement  because it means that the outstanding results can be used on virtually any topic without having to go through the expense and time of intensely training a model on specific knowledge domains.

What Medprompt Means For Generative AI

Medprompt has revealed a new way to elicit enhanced model capabilities, making generative AI more adaptable and versatile across a range of knowledge domains for a lot less training and effort than previously understood.

The implications for the future of generative AI are profound, not to mention how this may influence the skill of prompt engineering.

Read the new research paper:

Can Generalist Foundation Models Outcompete Special-Purpose Tuning? Case Study in Medicine (PDF)

Featured Image by Shutterstock/Asier Romero

ChatGPT: A Look Back At One Year Of AI Advances From OpenAI via @sejournal, @kristileilani

Today, ChatGPT celebrates one year since its launch in research preview.

From its humble beginnings, ChatGPT has continually pushed the boundaries of what we perceive as possible with generative AI for almost any task.

In this article, we take a journey through the past year, highlighting the significant milestones and updates that have shaped ChatGPT into the versatile and powerful tool it is today.

ChatGPT: From Research Preview To Customizable GPTs

This story unfolds over the course of nearly a year, beginning on November 30, when OpenAI announced the launch of its research preview of ChatGPT.

As users began to offer feedback, improvements began to arrive.

Before the holiday, on December 15, 2022, ChatGPT received general performance enhancements and new features for managing conversation history. To ensure quality, a daily message cap was also introduced.

As the calendar turned to January 9, 2023, ChatGPT saw improvements in factuality, and a notable feature was added to halt response generation mid-conversation, addressing user feedback and enhancing control.

Just a few weeks later, on January 30, the model was further upgraded for enhanced factuality and mathematical capabilities, broadening its scope of expertise.

February 2023 was a landmark month. On February 9, ChatGPT Plus was introduced, bringing new features and a faster ‘Turbo’ version to Plus users.

This was followed closely on February 13 with updates to the free plan’s performance and the international availability of ChatGPT Plus, featuring a faster version for Plus users.

March 14, 2023, marked a pivotal moment with the introduction of GPT-4 to ChatGPT Plus subscribers.

This new model featured advanced reasoning, complex instruction handling, and increased creativity.

Less than ten days later, on March 23, experimental AI plugins, including browsing and code interpreter capabilities, were made available to selected users.

On May 3, users gained the ability to turn off chat history and export data.

Plus users received early access to experimental web browsing and third-party plugins on May 12.

On May 24, the iOS app expanded to more countries with new features like shared links and Bing plugin, and an option to disable chat history on iOS.

June and July 2023 were filled with updates enhancing mobile app experiences and introducing new features.

The mobile app was updated with browsing features on June 22, and the browsing feature itself underwent temporary disablement for improvements on July 3.

The code interpreter feature rolled out in beta to Plus users on July 6.

Plus customers enjoyed doubled message limits for GPT-4 from July 19, and Custom instructions became available in beta to Plus users the next day.

July 25 saw the Android version of the ChatGPT app launch in selected countries.

As summer progressed, August 3 brought several small updates enhancing the user experience.

Custom instructions were extended to free users, except in the EU & UK, on August 9, and later to those regions on August 21.

The month concluded with the launch of ChatGPT Enterprise on August 28, offering advanced features and security for enterprise users.

Entering autumn, September 11 witnessed limited language support in the web interface.

Voice and image input capabilities in beta were introduced on September 25, further expanding ChatGPT’s interactive abilities.

An updated version of web browsing rolled out to Plus users on September 27.

The fourth quarter of 2023 began with integrating DALL·E 3 in beta on October 16, allowing for image generation from text prompts.

The browsing feature moved out of beta for Plus and Enterprise users on October 17. Customizable versions of ChatGPT, called GPTs, were introduced for specific tasks on November 6.

Finally, on November 21, the voice feature in ChatGPT was made available to all users, rounding off a year of significant advancements and broadening the horizons of AI interaction.

Looking Ahead: What’s Next For ChatGPT

The past year has been a testament to continuous innovation, but it is merely the prologue to a future rich with potential.

The upcoming year promises incremental improvements and leaps in AI capabilities, user experience, and integrative technologies that could redefine our interaction with digital assistants.

With a community of users and developers growing stronger and more diverse, the evolution of ChatGPT is poised to surpass expectations and challenge the boundaries of today’s AI landscape.

As we step into this next chapter, the possibilities are as limitless as generative AI continues to advance.

Screenshot from ChatGPT, November 2023

Featured image: photosince/Shutterstock

OpenAI Welcomes Back Sam Altman As CEO via @sejournal, @kristileilani

OpenAI has announced that Sam Altman is officially back as CEO.

Sam Altman is back as CEO, Mira Murati as CTO and Greg Brockman as President. OpenAI has a new initial board. Messages from @sama and board chair @btaylor https://t.co/sP0kAQIeKg

— OpenAI (@OpenAI) November 30, 2023

Sam Altman has returned as CEO, with Mira Murati reinstated as Chief Technology Officer (CTO).

The best interests of the company and the mission always come first. It is clear that there were real misunderstandings between me and members of the board. For my part, it is incredibly important to learn from this experience and apply those learnings as we move forward as a…

— Sam Altman (@sama) November 30, 2023

This leadership reshuffle coincides with the formation of a new initial board, a strategic move that underscores OpenAI’s commitment to its pioneering mission in artificial intelligence (AI).

The OpenAI team is irreplaceable — I couldn’t be happier to be back at work alongside @sama and @gdb . The mission continues.

— Mira Murati (@miramurati) November 30, 2023

The newly formed board includes Bret Taylor as Chair, alongside Larry Summers and Adam D’Angelo, bringing diverse expertise to guide OpenAI’s future endeavors.

Today as Board Chair I sent the attached letter on behalf of the Board to the OpenAI community. I am excited to be working with my board colleagues Adam and Larry, the OpenAI Leadership, and all OpenAI stakeholders. The Board’s focus is stabilizing the company, building out a…

— Bret Taylor (@btaylor) November 30, 2023

In his message to the company, Sam Altman emphasized the importance of resilience and the team’s hard work in navigating recent challenges.

Screenshot from X, November 2023

He expressed gratitude towards team members and partners, particularly highlighting the significant role of Microsoft.

Altman’s return signals a renewed focus on advancing AI research and safety initiatives, aligning with OpenAI’s long-term vision.

His leadership, coupled with Murati’s technological expertise and Greg Brockman’s continued role as President, sets a strong foundation for the company’s future growth.

Key priorities for the new leadership include enhancing research capabilities, deploying innovative products, and strengthening corporate governance.

These efforts aim to further OpenAI’s mission of ensuring that AI benefits humanity while addressing the evolving needs of users and stakeholders.

For marketers and SEO professionals, OpenAI’s developments hold particular significance.

The company’s technologies, like ChatGPT, have already transformed marketing strategies and search engine optimization.

Continued advancements in these areas promise new tools and insights for the industry.

In summary, OpenAI’s leadership transition with Sam Altman at the helm, a new board, and a clear focus on research and governance, heralds an exciting era for the AI community.

This shift is poised to impact various sectors, offering fresh perspectives and innovative solutions in the world of AI and technology.

Featured image: Patrickx007/Shutterstock