🧐The curious case of gpt2

In partnership with

Reading time: 5 minutes  

 Newsletter | News | Tools | Submit Tool  

Hire a world class AI team

Engineers who understand AI are expensive and difficult to find, and it can be hard to figure out who to trust. On top of that, 85% of all AI projects fail.

But AE Studio succeeds.

We listen to your business challenge and help you craft and implement the optimal AI solution with our team of world class AI experts from Harvard, Stanford and Princeton.

Our development, design, and data science teams work closely with founders and executives to create custom software and AI solutions that get the job done. The secret to our success is treating your project as if it were our own startup.

Tell us all about your big AI Idea

Today we will discuss-

• 🦄All about the mysterious gpt2-chatbot

• ✊🏻Austria pushes for regulation of lethal AI

• 🎯NIST targets AI misinformation surge

• ⚙️9 amazing AI tools you might not have heard of

All this and more - Let's dive in!

👩‍🍳What’s cooking in the newsroom?

Mysterious gpt2-chatbot shakes up the AI community

♨️News - A powerful new AI system surfaced on the internet today, sparking speculation about its origins and capabilities, with some researchers suggesting it represents a significant leap forward compared to existing AI models.

🤹🏻‍♂️Here's everything you need to know -

• The model, named "gpt2-chatbot," made its debut on LMSYS Chatbot Arena, a website known for comparing AI language systems, without any prior announcement or promotion. 

• Andrew Gao, a Stanford University student and AI researcher, tested the chatbot's mathematical abilities by presenting it with an International Math Olympiad problem. Impressively, the chatbot solved it on its first attempt.

• According to Ethan Mollick, a professor at the Wharton School of the University of Pennsylvania who studies AI, in his experiments, the model outperformed GPT-4 on complex reasoning tasks, such as writing code to create a picture of a unicorn.

• Meanwhile, AI influencer Rowan Cheung praised the chatbot's proficiency in ASCII art, describing it as "miles ahead of any other model" in this area.

• Furthermore, Chase, a founding engineer at Codegen, stated that the chatbot's coding skills surpassed those of GPT-4 and Claude Opus. He mentioned that the gpt2-chatbot performed exceptionally well in complex code manipulation tasks, even outperforming newer models.

• Several users on X pointed out that the model identifies itself as "ChatGPT, a large language model trained by OpenAI, based on the GPT-4 architecture." Speculation that the model is the creation of OpenAI grew after Sam Altman, the CEO and co-founder of OpenAI, mentioned, "I do have a soft spot for gpt2" in a post on X.

• However, it's still possible that "gpt2-chatbot" could have originated from a lesser-known company or research group aiming to showcase their AI capabilities and generate attention. Regardless, it shows just how fast AI is evolving and how tough it is to keep up with all the latest developments!

Austria urges global regulation of autonomous weapons systems amidst 'killer robots' concerns

💣Context of the news - With AI technology advancing rapidly, the development of weapons systems capable of autonomous killing is becoming more imminent. This presents ethical and legal challenges that most countries agree need to be addressed soon.

☕News - On Monday, Austria urged for renewed efforts to regulate the use of artificial intelligence in weapons systems, which could potentially lead to the creation of 'killer robots'. This call came during a conference hosted by Austria, aimed at reviving discussions on the issue, which have largely stalled.

🎤Key quotes - 

• Addressing representatives from non-governmental and international organizations, as well as envoys from 143 countries, Austrian Foreign Minister Alexander Schallenberg urged immediate action, stating, "We cannot let this moment pass without taking action. Now is the time to establish international rules and norms to ensure human control."

• "At least let us make sure that the most profound and far-reaching decision— who lives and who dies—remains in the hands of humans and not machines," he added.

• During a panel discussion at the conference, Mirjana Spoljaric, President of the International Committee of the Red Cross, warned against accelerating moral failures by entrusting control of violence to machines and algorithms, stating, "What we see today in different contexts of violence are moral failures in the face of the international community. And we do not want to see such failures accelerating by giving the responsibility for violence, for the control over violence, over to machines and algorithms."

🥸In conclusion - Sadly, years of discussions at the United Nations have resulted in few tangible results, and many participants at the two-day conference in Vienna agreed that the window for action was rapidly closing. What's your take? 🤷🏻‍♂️

NIST launches new initiative to combat rising tide of AI-generated misinformation

👨🏻‍💻News - The National Institute of Standards and Technology (NIST), a branch of the U.S. Commerce Department that develops and tests tech for the U.S. government, businesses, and the public, has introduced a new program called NIST GenAI. 

🕵🏻‍♂️What's that? This new initiative is aimed at evaluating generative AI technologies, such as those used for generating text and images. 

The program will introduce benchmarks and support the development of content authenticity detection systems, including those for detecting deepfakes. Additionally, it will encourage the creation of software capable of identifying the origin of fake or misleading AI-generated information. 

In line with this, NIST GenAI's first project is a pilot study focused on building systems that can effectively differentiate between human-created and AI-generated media, starting with text.

🤠Why is this important? While the initiative spearheaded by NIST is in response to President Joe Biden's executive order on AI, the launch of NIST GenAI, particularly its focus on deepfakes, is worth watching, given the exponential growth in the volume of AI-generated deceptive content. According to Clarity, a deepfake detection firm, there has been a 900% increase in the creation and publication of deepfakes this year compared to the same period last year.

🙆🏻‍♀️What else is happening?

• ChatGPT’s ‘hallucination’ problem hit with another privacy complaint in EU

• Google sued by US artists over AI image generator

• Second global AI safety summit faces tough questions, lower turnout

• OpenAI to use FT content for training AI models in latest media tie-up

• GitHub Copilot can now help start a project with AI, not just complete it

👨🏻‍🎓Prompt engineering tips

• Provide Constraints - If necessary, provide constraints or guidelines to help the model generate appropriate responses. Set boundaries to ensure the model stays on topic and provides relevant information. Example: "Write a haiku about the changing seasons. Remember to follow the 5-7-5 syllable pattern."

• Provide Feedback - Give feedback to the model based on its responses. Let the model know what it did well and where it can improve. Example: "Good job summarizing the main points, but could you try again and be more concise this time?"

• Avoid Ambiguity - Use clear and unambiguous language. Avoid vague terms or phrasing that could be interpreted in multiple ways. Example: Instead of saying: "Summarize the key points of the research paper in one paragraph." Say: "Summarize the key points of the research paper in one hundred words."

👩🏼‍🚒Discover mind-blowing AI tools

1. Virtual CMO - A tool that helps solopreneurs solve any marketing problem in just one minute (Free)

2. Mixpanel Spark - A tool that allows users to ask questions in natural language and get answers about their product, marketing, and revenue data ($24/month)

3. KrockIO - A one-stop solution for remote creative collaboration designed for animation and video production firms ($15/month)

4. Adori Labs - A platform that helps bloggers convert their written content into engaging videos using AI (9.99/month) 

5. Watermelon - An all-in-one automated customer service tool powered by GPT-4 ($100/month)

6. Typeboss - It is a user-friendly AI copywriting tool, powered by OpenAI's GPT 3.5 model (Free, $2.99/month) 

7. Wepik - An online platform that offers various design templates and tools to create professional graphics (Free)

8. AINiro - An AI chatbot platform that offers custom ChatGPT chatbots for various purposes such as customer service, e-commerce, and lead generation ($300/month) 

9. Keyframes Studio - An all-in-one platform for creating, editing, and repurposing videos for social media platforms ($9/month)

🎨Unusual AI use cases

• Movie Trailers - AI algorithms analyze movie footage, soundtracks, and audience reactions to create movie trailers. For example, "IBM Watson" created the world's first AI-generated movie trailer for the horror film "Morgan," using machine learning algorithms to analyze the film's scenes and identify the most suspenseful moments.

• Art Restoration - AI algorithms are used to restore and enhance old artworks and photographs. For example, the "Refik Anadol Studio" used AI to restore and colorize old photographs of the historic Hagia Sophia Museum in Istanbul, bringing them to life with vivid colors and details.

• Compose Poetry - AI algorithms, such as OpenAI's GPT-3, can compose poetry in various styles. For example, "Sudowrite" is an AI-powered writing assistant that can generate poetry based on user prompts, mimicking the style of famous poets or creating original works.