- FryAI
- Posts
- Conversational search is here
Conversational search is here
Good morning! Welcome to a new week of AI insights. All weekend, we have been cooking up the latest just for you. 😁
🤯 MYSTERY AI LINK 🤯
(The mystery link can lead to ANYTHING AI-related: tools, memes, articles, videos, and more…)
Today’s Menu
Appetizer: OpenAI makes internet search conversational 💬
Entrée: Google’s code is 25% written by AI 🤖
Dessert: OpenAI introduces a way to “fact check” AI models 👍
🔨 AI TOOLS OF THE DAY
🤝 Hiring Studio: Generate interview questions using AI. → Check it out
📕 Nextbook: Find your next book with the help of an AI librarian. → Check it out
🍿 MovieToEmoji: Convert movie names into emojis. → Check it out
OPENAI MAKES INTERNET SEARCH CONVERSATIONAL 💬
search is my favorite feature we have launched in chatgpt since the original the launch!
it has probably doubled my usage over the past few weeks.
— Sam Altman (@sama)
5:32 PM • Oct 31, 2024
Searching the internet will never be the same! 🛜
What’s up? OpenAI has officially launched ChatGPT Search, a new internet search feature within the ChatGPT platform.
How does it work? ChatGPT Search, initially beta-tested in July, is designed to deliver real-time updates on sports scores, stock quotes, news, weather, and more. The new feature is powered by live web searches and partnerships with key data and news providers, making it a significant shift in how users can gather information. Users can easily access links to original sources directly from responses.
Why is this significant? CEO Sam Altman described the search capability as his “favorite feature” since ChatGPT’s debut, highlighting its potential to simplify complex research and offer seamless follow-up questioning. This move positions OpenAI to compete more directly with major search engines like Google, as well as with AI-driven platforms such as Perplexity. Regardless of whether ChatGPT Search replaces Google as a leader in the space, one thing is for certain: web search is getting more conversational. As Altman points out, “It’s hard to go back to doing it the old way.”
GOOGLE’S CODE IS 25% WRITTEN BY AI 🤖
My computer said my password is insecure … Maybe if it didn’t have such strict requirements, it would be more confident. 👾
What’s new? During Google’s third-quarter earnings call, CEO Sundar Pichai revealed that more than a quarter of Google’s new code is now generated by AI.
“We’re using AI internally to improve our coding processes, which is boosting productivity and efficiency. Today, more than a quarter of all new code at Google is generated by AI, then reviewed and accepted by engineers. This helps our engineers do more and move faster.”
Want the details? While this may seem like a cause for concern about the replacement of human programmers, Google views it quite differently. This internal use of AI has significantly boosted productivity and efficiency, allowing engineers to accelerate their work. Pichai also emphasized the role of humans in reviewing this code, underscoring their importance. This innovation highlights the transformative role of AI in software development as Google stays competitive in the tech industry.
OPENAI INTRODUCES A WAY TO “FACT CHECK” AI MODELS 👍
How do we fact check the facts used by OpenAI to fact check? 🙃
What’s new? OpenAI has introduced a new open-source benchmark, called SimpleQA, to measure how factual an AI model is.
What’s the problem? An ongoing challenge in AI is training models to produce factually correct responses. Language models, while impressive, often generate incorrect or unsupported outputs—a problem known as “hallucinations.” Addressing this issue is crucial for building trustworthy AI systems.
How does this tool help solve the problem? SimpleQA is designed to assess AI models through short, fact-seeking queries, making evaluation more manageable. Each question in SimpleQA has a single, clear answer verified by two independent AI trainers. Questions span a wide array of topics, from science and technology to video games and politics, providing diversity and challenging even advanced models. A third AI trainer validated a sample of these questions, achieving a 94.4% agreement rate, underscoring the benchmark’s reliability. By grading responses as “correct,” “incorrect,” or “not attempted,” SimpleQA helps researchers measure and compare models’ accuracy and calibration, promoting further progress toward more factual and dependable AI.
MONEY MONDAY 🤑
People are discovering innovative (and sometimes wacky) ways to make money using AI. Check out today’s featured video:
HAS AI REACHED SINGULARITY? CHECK OUT THE FRY METER BELOW:
What do ya think of this latest newsletter? |