March 15, 2023

gpt-4

5 jaw-dropping things GPT-4 can do that ChatGPT couldn’t

CNN, 03/15/23. OpenAI’s release of GPT-4, the next-generation version of its large language model. It has stunned users by working with photos, passing standardized exams, and creating functional websites from hand-drawn sketches. 

This update promises to change the way we use the internet to work, play, create and use the Internet. It amplifies the challenging questions around how AI tools can upend professions and enable cheating. 

GPT-4 has the potential to streamline work across most industries. Users are already generating “one click lawsuits” and better match dating sites. This new technology still has its limitations. Users are cautioned to exercise great care particularly when the stakes are high. READ MORE

healthcare

Google flexes its health care AI muscle

AXIOS, 03/15/23. Google has introduced AI healthcare tools including the Med-PaLM 2 chatbot, Duplex AI for Medicaid acceptance, and AI-powered diagnostics tools. 

Google is addressing concerns about AI in healthcare by adding information panels to YouTube Health and partnering with ThroughLine for free mental health crisis support. 

AI holds immense potential for improving public health. But nothing will beat the human touch. READ MORE

responsiblE AI

Microsoft picks perfect time to dump its AI ethics team

The Register, 03/15/23. Microsoft dumped its AI ethics team during its recent round of layoffs. This maneuver has generated concern about the company’s commitment to responsible AI practices. 

The ethics team had a crucial role in advising product teams on ethical standards, and its removal may undermine vital guardrails. Microsoft says it maintains a commitment to responsible AI through other teams. 

The company is investing heavily in OpenAI and integrating large language model AI systems across its product lines. Thus, it’s critical to have ongoing scrutiny and discussions on AI’s impact on society. READ MORE

limitations

I Had AI Write My Instagram Captions, And My Friends Did A Wellness Check On Me

BuzzFeedNews, 03/15/23. In a recent experiment, a journalist used an AI writing tool to generate Instagram captions for two weeks. She was able to generate AI text in response to prompts or shortened versions of desired text. 

While the AI-generated captions were legible and understandable, they lacked the personal tone and nuance that is typically associated with Instagram captioning. This led some of her followers to question the authenticity of the posts. 

Ultimately, the experiment revealed more about the journalist’s own character flaws than the capabilities. She did not recommend any one else give it a try.. READ MORE

integration

The Nightmare of AI-Powered Gmail Has Arrived

New York Magazine, 03/15/23. Google is playing catch-up in the development of AI-powered tools. It’s bolting on AI capabilities to Gmail, Google Docs, and a chatty version of its search engine. 

These features may solve one set of user problems but they risk creating others. The institutional value of some tasks could shrink while the value of others expands. 

The change represents just a glimpse of what might come after the AI hype dies down. Users should prepare themselves for genuinely bracing technologies to come. READ MORE