NEWS

An interview with Dr. Stuart Russell, author of “Human Compatible, Artificial Intelligence and the Problem of Control”

Published

5 years ago

October 6, 2019

an interview with dr stuart russell author of human compatible artificial intelligence and the problem of control

(UC Berkeley’s Dr. Stuart Russell’s new book, “Human Compatible: Artificial Intelligence and the Problem of Control, goes on sale Oct. 8. I’ve written a review, “Human Compatible” is a provocative prescription to re-think AI before it’s too late,” and the following in an interview I conducted with Dr. Russell in his UC Berkeley office on September 3, 2019.)

Ned Desmond: Why did you write Human Compatible?

Dr. Russell: I’ve been thinking about this problem – what if we succeed with AI? – on and off since the early 90s. The more I thought about it, the more I saw that the path we were on doesn’t end well.

(AI Researchers) had mostly just doing toy stuff in the lab, or games, none of which represented any threat to anyone. It’s a little like a physicist playing tiny bits of uranium. Nothing happens, right? So we’ll just make more of it, and everything will be fine. But it just doesn’t work that way. When you start crossing over to systems that are more intelligent, operating on a global scale, and having real-world impact, like trading algorithms, for example, or social media content selection, then all of a sudden, you are having a big impact on real-world, and it’s hard to control. It’s hard to undo. And that’s just going to get worse and worse and worse.

Stuart Russell HUMAN COMPATIBLE Credit Peg Skorpinski

Dean’s Society – October 23, 2006; Stuart Russell

Desmond: Who should read Human Compatible?

Dr. Russell: I think everyone, because everyone is going to be affected by this. As progress occurs towards human level (AI), each big step is going to magnify the impact by another factor of 10, or another factor of 100. Everyone’s life is going to be radically affected by this. People need to understand it. More specifically, it would be policymakers, the people who run the large companies like Google and Amazon, and people in AI, related disciplines, like control theory, cognitive science and so on.

My basic view was so much of this debate is going on without any understanding of what AI is. It’s just this magic potion that will make things intelligent. And in these debates, people don’t understand the building blocks, how it fits together, how it works, how you make an intelligent system. So chapter two (of Human Compatible was) sort of mammoth and some people said, “Oh, this is too much to get through and others said, “No, you absolutely have to keep it.” So I compromised and put the pedagogical stuff in the appendices.

Desmond: Why did computer scientists tend to overlook the issue of uncertainty in the objective function for AI systems?

Dr. Russell: Funnily enough, in AI, we took uncertainty (in the decision-making function) to heart starting in the 80s. Before that, most AI people said let’s just work on cases where we have definite knowledge, and we can come up with guaranteed plans.

Up Next

Mobile conversions rising, but desktop conversions still 93% more valuable, study says

Don't Miss

PayPal is the first company to drop out of the Facebook-led Libra Association

NEWS

OpenAI Introduces Fine-Tuning for GPT-4 and Enabling Customized AI Models

Published

3 weeks ago

August 21, 2024

Max

OpenAI Introduces Fine-Tuning for GPT-4 and Enabling Customized AI Models

OpenAI has today announced the release of fine-tuning capabilities for its flagship GPT-4 large language model, marking a significant milestone in the AI landscape. This new functionality empowers developers to create tailored versions of GPT-4 to suit specialized use cases, enhancing the model’s utility across various industries.

Fine-tuning has long been a desired feature for developers who require more control over AI behavior, and with this update, OpenAI delivers on that demand. The ability to fine-tune GPT-4 allows businesses and developers to refine the model’s responses to better align with specific requirements, whether for customer service, content generation, technical support, or other unique applications.

Why Fine-Tuning Matters

GPT-4 is a very flexible model that can handle many different tasks. However, some businesses and developers need more specialized AI that matches their specific language, style, and needs. Fine-tuning helps with this by letting them adjust GPT-4 using custom data. For example, companies can train a fine-tuned model to keep a consistent brand tone or focus on industry-specific language.

Fine-tuning also offers improvements in areas like response accuracy and context comprehension. For use cases where nuanced understanding or specialized knowledge is crucial, this can be a game-changer. Models can be taught to better grasp intricate details, improving their effectiveness in sectors such as legal analysis, medical advice, or technical writing.

Key Features of GPT-4 Fine-Tuning

The fine-tuning process leverages OpenAI’s established tools, but now it is optimized for GPT-4’s advanced architecture. Notable features include:

Enhanced Customization: Developers can precisely influence the model’s behavior and knowledge base.
Consistency in Output: Fine-tuned models can be made to maintain consistent formatting, tone, or responses, essential for professional applications.
Higher Efficiency: Compared to training models from scratch, fine-tuning GPT-4 allows organizations to deploy sophisticated AI with reduced time and computational cost.

Additionally, OpenAI has emphasized ease of use with this feature. The fine-tuning workflow is designed to be accessible even to teams with limited AI experience, reducing barriers to customization. For more advanced users, OpenAI provides granular control options to achieve highly specialized outputs.

Implications for the Future

The launch of fine-tuning capabilities for GPT-4 signals a broader shift toward more user-centric AI development. As businesses increasingly adopt AI, the demand for models that can cater to specific business needs, without compromising on performance, will continue to grow. OpenAI’s move positions GPT-4 as a flexible and adaptable tool that can be refined to deliver optimal value in any given scenario.

By offering fine-tuning, OpenAI not only enhances GPT-4’s appeal but also reinforces the model’s role as a leading AI solution across diverse sectors. From startups seeking to automate niche tasks to large enterprises looking to scale intelligent systems, GPT-4’s fine-tuning capability provides a powerful resource for driving innovation.

OpenAI announced that fine-tuning GPT-4o will cost $25 for every million tokens used during training. After the model is set up, it will cost $3.75 per million input tokens and $15 per million output tokens. To help developers get started, OpenAI is offering 1 million free training tokens per day for GPT-4o and 2 million free tokens per day for GPT-4o mini until September 23. This makes it easier for developers to try out the fine-tuning service.

As AI continues to evolve, OpenAI’s focus on customization and adaptability with GPT-4 represents a critical step in making advanced AI accessible, scalable, and more aligned with real-world applications. This new capability is expected to accelerate the adoption of AI across industries, creating a new wave of AI-driven solutions tailored to specific challenges and opportunities.

GOOGLE

This Week in Search News: Simple and Easy-to-Read Update

Published

3 months ago

June 30, 2024

Entireweb News Bot

This Week in Search News: Simple and Easy-to-Read Update

Here’s what happened in the world of Google and search engines this week:

1. Google’s June 2024 Spam Update

Google finished rolling out its June 2024 spam update over a period of seven days. This update aims to reduce spammy content in search results.

2. Changes to Google Search Interface

Google has removed the continuous scroll feature for search results. Instead, it’s back to the old system of pages.

3. New Features and Tests

Link Cards: Google is testing link cards at the top of AI-generated overviews.
Health Overviews: There are more AI-generated health overviews showing up in search results.
Local Panels: Google is testing AI overviews in local information panels.

4. Search Rankings and Quality

Improving Rankings: Google said it can improve its search ranking system but will only do so on a large scale.
Measuring Quality: Google’s Elizabeth Tucker shared how they measure search quality.

5. Advice for Content Creators

Brand Names in Reviews: Google advises not to avoid mentioning brand names in review content.
Fixing 404 Pages: Google explained when it’s important to fix 404 error pages.

6. New Search Features in Google Chrome

Google Chrome for mobile devices has added several new search features to enhance user experience.

7. New Tests and Features in Google Search

Credit Card Widget: Google is testing a new widget for credit card information in search results.
Sliding Search Results: When making a new search query, the results might slide to the right.

8. Bing’s New Feature

Bing is now using AI to write “People Also Ask” questions in search results.

9. Local Search Ranking Factors

Menu items and popular times might be factors that influence local search rankings on Google.

10. Google Ads Updates

Query Matching and Brand Controls: Google Ads updated its query matching and brand controls, and advertisers are happy with these changes.
Lead Credits: Google will automate lead credits for Local Service Ads. Google says this is a good change, but some advertisers are worried.
tROAS Insights Box: Google Ads is testing a new insights box for tROAS (Target Return on Ad Spend) in Performance Max and Standard Shopping campaigns.
WordPress Tag Code: There is a new conversion code for Google Ads on WordPress sites.

These updates highlight how Google and other search engines are continuously evolving to improve user experience and provide better advertising tools.

FACEBOOK

Facebook Faces Yet Another Outage: Platform Encounters Technical Issues Again

Published

6 months ago

March 20, 2024

Max

Uppdated: It seems that today’s issues with Facebook haven’t affected as many users as the last time. A smaller group of people appears to be impacted this time around, which is a relief compared to the larger incident before. Nevertheless, it’s still frustrating for those affected, and hopefully, the issues will be resolved soon by the Facebook team.

Facebook had another problem today (March 20, 2024). According to Downdetector, a website that shows when other websites are not working, many people had trouble using Facebook.

This isn’t the first time Facebook has had issues. Just a little while ago, there was another problem that stopped people from using the site. Today, when people tried to use Facebook, it didn’t work like it should. People couldn’t see their friends’ posts, and sometimes the website wouldn’t even load.

Downdetector, which watches out for problems on websites, showed that lots of people were having trouble with Facebook. People from all over the world said they couldn’t use the site, and they were not happy about it.

When websites like Facebook have problems, it affects a lot of people. It’s not just about not being able to see posts or chat with friends. It can also impact businesses that use Facebook to reach customers.

Since Facebook owns Messenger and Instagram, the problems with Facebook also meant that people had trouble using these apps. It made the situation even more frustrating for many users, who rely on these apps to stay connected with others.

During this recent problem, one thing is obvious: the internet is always changing, and even big websites like Facebook can have problems. While people wait for Facebook to fix the issue, it shows us how easily things online can go wrong. It’s a good reminder that we should have backup plans for staying connected online, just in case something like this happens again.