Daily Dose of Tech | 2023-12-19

OpenAI buffs safety team and gives board veto power on risky AI, Midjourney alpha released, 8 predictions for AI in 2024

Hello Tech Enthusiasts! ??

Welcome to Edition 12 of my Daily Dose of Tech, your hub for all things Tech, Cloud Computing & AI. Stay updated with the latest news, expert insights, practical tips, and spotlights on emerging tech trends. Let's dive right in!

?? OpenAI buffs safety team and gives board veto power on risky AI

OpenAI is enhancing its safety measures against potential harms of AI technology by forming a new "safety advisory group" that will oversee its technical teams and provide recommendations to the company's leadership, with the board having veto power. This development comes amidst a broader discussion on AI risks and follows recent changes in the company's leadership, including the departure of key board members. The focus is on how OpenAI, a leading AI development company, is handling safety considerations in the wake of these changes.

OpenAI's updated "Preparedness Framework" outlines a structured approach to identifying and mitigating catastrophic risks associated with their AI models. These risks are defined as those causing significant economic damage or harm to individuals, including existential threats. The framework categorizes risks into cybersecurity, persuasion (like disinformation), model autonomy, and CBRN (chemical, biological, radiological, and nuclear) threats. High-risk models cannot be deployed, and those with critical risks are halted in development. Additionally, the company's decision-making process includes the simultaneous involvement of both the board and leadership, with the board having the power to reverse decisions. However, questions remain about the transparency and efficacy of this new structure, especially considering the recent changes in board composition and the potential conflicts between expert recommendations and board decisions.

Read the full article here ??

??? Midjourney alpha released

Midjourney, one of the most popular image-generating AI services, announced the launch of the “Alpha” version of its website, available to a select few users.

Unfortunatly the new site is currently only available to those who have amassed over 10,000 images through Midjourney on Discord. But the tech company confirmed that the website will roll out to the rest of us soon. I'll let you know when the new page is publicly accessible.

Under the link you can see pictures which have been created with Midjourney:

https://legacy.midjourney.com/showcase/recent/

8 predictions for AI in 2024 from TechCrunch

• OpenAI becomes a product company: After leadership changes, OpenAI is shifting towards a product-focused approach, evident in the planned launch of the GPT store, an "app store for AI." This shift marks a departure from the company's previous caution and academic focus, moving towards a market-driven approach.
• Agents, generated video, and music: In 2024, AI applications in agent-based models and generative multimedia are expected to evolve from experimental to more practical uses. These include agent models assisting in tasks like insurance claims and generative music being used in professional settings like video games.
• Limits of Large Language Models (LLMs): The capabilities and limits of LLMs will become more apparent in 2024, with a better understanding of their shortcomings and a focus on diminishing returns in terms of model size and compute requirements. There will be a shift towards a mix of smaller, specialized models for efficiency.
• Marketing meets reality: The hype around AI in 2023 may lead to customer disillusionment in 2024 as promised benefits fail to materialize, potentially resulting in lawsuits and regulatory actions against AI service providers. This could lead to a market consolidation as weaker AI products and companies are weeded out.
• Apple jumps into AI: Apple is expected to enter the AI space with a focus on practical applications, using its ecosystem to integrate AI in everyday tasks. They might either improve Siri or launch a new AI service, leveraging their track record of refining and polishing technology.
• Legal cases in AI: Numerous AI-related lawsuits filed in 2023 are expected to either stall or lead to settlements in 2024, with new lawsuits emerging around AI misuse in areas like hiring and lending. However, the lack of specific AI laws may limit court actions.
• Early adopters of AI regulations: The EU's AI Act and similar regulations will slowly change the AI industry, leading to the growth of an AI compliance sector. However, substantial AI regulation in the U.S. is unlikely in 2024 due to the focus on the presidential election.
• The 2024 election and AI's role: AI is expected to significantly influence the 2024 presidential election, with AI-generated content like bot accounts and fake news being used for manipulation. Efforts to control such content are likely to be insufficient, exacerbating misinformation and distrust in the electoral process.

Read the full article here ??

Hey there, I'm Tobias Schümann, the visionary behind Cloud Consulting Europe.

I'm passionate about reshaping industries using cutting-edge cloud tech, AI, and innovation. Throughout Europe, I've led various projects, from smooth cloud migrations to game-changing IoT solutions.

Bagging all 12 AWS certifications shows how deep I'm into mastering the cloud. I love discussing digitization and the ever-changing world of technology.

Let's chat and explore the endless possibilities of tech together or simply let me know about other tech topics I should cover in the future. Gladly also by mail: [email protected].

Tech is for everyone ??