How I create this newsletter with the help of AI
How is the Smart Content Report actually created? First of all, I have to keep myself up to date. I use NetNewsWire as my feed reader. It allows me to keep track of new posts on dozens of websites in one central place.
For the Smart Content Report, three sources are currently the most important:
I bookmark what I find in these sources in my browser. I already sort them into their respective categories (news, tools, background, etc.).
Until now, I have been going through these links every 14 days, sorting them and processing them for the newsletter. This has changed, because the Smart Content Report now has its own website . Several times a week, I’ll post what I’ve found on that site and use these posts to create this newsletter.
I will describe in more detail why I started the site another time.
Back to my process: All posts are created with the help of AI. My tool of choice is usually Claude 3.5 Sonnet, as it has the best language capabilities.
I have two prompts for Claude. Quick heads-up: The originals are in German and I translated them for this post. Therefore, they might not have the same effect in English.
A major difference with the second prompt is that I tell the AI to first create an “overview of the main points and statements”. In my experience, this helps to improve the quality of the actual text.
In both cases, I review the results and improve the wording and other details. If something is unclear to me, I ask the AI for an explanation and possibly a new version. And of course I check the statements and details.
I do all this in German first, as I have more experience with writing and editing in German.
I then use DeepL Translator to translate my German texts into English and DeepL Write to improve the style of the automatic translation.
I find the results pretty good and I rarely have to correct or change much if anything.
The newsletter is ultimately published in German by UPLOAD Magazin and I send it out in English via LinkedIn . Eventually I would like to have the newsletter on the Smart Content Report's website instead. It might also make sense to send it every week instead of every two weeks. But I’m still deciding.
That’s my process in a nutshell. If you have any questions or feedback, please feel free to leave a comment!
T O O L S
OpenAI o1 impresses with surprisingly strong performance for some tasks
OpenAI has unveiled a new family of AI models called “o1”. It was previously known as “Project Strawberry” and had led to all kinds of speculation and high expectations.
The first two versions, o1-preview and o1-mini, use a reasoning method known as “chain of thought” to solve complex tasks. This technique allows the models to think longer before generating answers and to approach problems step-by-step, much like humans would.
According to OpenAI, the new models show significantly improved performance in areas such as mathematics, programming, and science. In benchmarks, o1-preview achieved PhD-level performance in some scientific disciplines and significantly outperformed GPT-4 on Math Olympiad tasks. In programming competitions on Codeforces, the model reached the 89th percentile of participants.
The development of o1 is based on a novel training method using reinforcement learning. The model is trained using rewards and punishments, resulting in improved problem-solving ability. OpenAI states that o1 is less prone to hallucinations than previous models, but emphasizes that the problem is not completely solved.
Despite the advances, o1 also has limitations. The models are slower than GPT-4 and significantly more expensive in terms of API usage. In addition, features such as web search or image analysis are still missing in the early stages. Availability is initially limited to ChatGPT Plus and Team users, with plans to expand to other user groups.
OpenAI emphasizes the focus on safe and ethical use of the new models. The company has entered into agreements with US and UK AI safety institutes and is conducting extensive internal testing. OpenAI plans to continue to develop the o1 model family, adding more features to increase its usefulness and accessibility for various applications.
Sources: OpenAI ,?The Verge ,?TechCrunch ,?VentureBeat
Salesforce unveils Agentforce: AI Agents for customer service and more
Salesforce has now officially unveiled Agentforce, a previously announced suite of AI-powered autonomous agents. These are designed to assist human employees in various business areas such as sales, marketing, commerce, and customer service. According to Salesforce CEO Marc Benioff, Agentforce represents the “third wave” of AI, moving beyond trending models and generative AI.
At the heart of Agentforce is Atlas, a new reasoning engine that Salesforce says is designed to simulate human thinking and planning. The platform offers pre-built agents for different roles, as well as low-code options for customization. Early adopters report significant productivity gains. Educational publisher Wiley, for example, has seen a self-service resolution rate of more than 40 percent for customer inquiries.
The launch of Agentforce comes at a time of growing concern about the impact of AI on jobs and privacy. Salesforce emphasizes its commitment to the ethical use of AI and says it has built in appropriate safeguards.
Salesforce plans to unveil Agentforce at its annual Dreamforce conference and sign up at least 1,000 customers for the new platform.
Sources: Axios , VentureBeat
Microsoft shows off Copilot Pages, other AI features for businesses
Microsoft is expanding its Office offerings with new AI features. A key innovation is “Copilot Pages”, a platform for collaborative work with AI. Users can work together on AI-generated content and enrich it with data from different sources. Microsoft describes this as an entirely new way of working that brings humans and AI together in a shared workspace.
Microsoft is also introducing enhancements to Excel, Outlook, and other Office applications. These include AI-powered charting capabilities in Excel, email prioritization in Outlook, and the introduction of AI agents that can act as virtual employees and automate routine tasks.
At the same time, Microsoft won a major new customer in Vodafone. The telecommunications company purchased 68,000 licenses for the Copilot AI assistants after tests showed time savings of about three hours per week per employee.
Snapchat introduces AI video tool
Snapchat is introducing a new AI-powered video creation tool. The tool, which is now in beta for select users, allows users to create videos from text prompts, and soon, image prompts. The company made the announcement at its annual Snap Partner Summit, as Aisha Malik reports for TechCrunch.
Snap stresses that all content created with the tool will have a visible watermark. The company also assures that the underlying AI models will undergo security assessments and testing to prevent harmful content. With this launch, Snap is positioning itself ahead of competitors like TikTok and Instagram, which have yet to release comparable text-to-video AI generators.
Pixtral 12B: Mistral’s first multimodal model
French AI startup Mistral has released its first multimodal model, Pixtral 12B. In other words, it has 12 billion parameters and can process both images and text. It is based on Mistral’s existing text model Nemo 12B and is said to be able to answer questions about any number of images of any size.
Pixtral 12B is currently freely available via GitHub and the Hugging Face AI development platform under an Apache 2.0 license. According to Mistral, the model can be downloaded, customized, and used without restrictions. A public demo version is not yet available, but will soon be accessible via Mistral’s chatbot and API platforms.
With this new model, Mistral is expanding its portfolio of AI models and competing directly with established providers such as OpenAI and Anthropic. The one-year-old startup, in which Microsoft holds a minority stake, is seen as the European answer to OpenAI and recently raised $645 million in a funding round. The company’s strategy so far includes releasing free open models, paid versions, and consulting services for enterprise customers.
Sources: TechCrunch ,?VentureBeat
DeepSeek V2.5 celebrated as new open source champion
DeepSeek-V2.5 is the new champion among open source AI models. DeepSeek itself is an offshoot of the Chinese hedge fund High-Flyer Capital Management. The new model combines natural language processing and programming capabilities into one seemingly powerful system. According to Carl Franzen of VentureBeat, DeepSeek-V2.5 outperforms its predecessors in almost all benchmarks. The model offers new features such as function calling and shows improved performance in writing, following instructions, and advanced coding. Independent researchers confirm the outstanding performance of the model, which is freely available under a modified MIT license on Hugging Face.
10 AI tools for copywriting
In a new article on the Hootsuite website, Sarah Israel presents what she believes are the “10 best AI tools for copywriting” in 2024. These tools can help with various tasks such as brainstorming, writing social media posts, or search engine optimization. Among the offerings featured are OwlyWriter AI, ChatGPT, and HubSpot Content Writer. The author emphasizes that AI tools should not replace human creativity, but rather complement it. She also recommends always reviewing and editing the generated content.
N E W S
OpenAI o1 “not a breakthrough towards AGI”
The ARC research team has tested OpenAI’s new AI models o1-preview and o1-mini with the ARC AGI benchmark . ARC stands for “Alignment Research Center” and their AGI benchmark measures the capabilities of AI systems in terms of Artificial General Intelligence (AGI). They define AGI as follows:
AGI is a system that can efficiently learn new skills and solve open-ended problems.
So far, all tested systems have been far from human capabilities in this area.
Although the o1 models did outperform GPT-4 in terms of accuracy in this test, they required significantly more computing time. They use a “chain-of-thought” approach in which the AI generates intermediate steps to refine and check its answer in advance. According to ARC, o1 shows progress, but does not represent a breakthrough towards AGI. The researchers see a need for further innovation to develop AI systems that go beyond simply applying learned patterns to truly generate new solutions.
OpenAI wants to shed its non-profit roots
OpenAI, the company behind ChatGPT, is on the verge of a major restructuring. CEO Sam Altman has apparently told employees that the startup plans to shed its complex non-profit structure. The change is expected to take place in 2025 and could turn OpenAI into a more traditional, for-profit company.
The planned restructuring coincides with a new round of funding in which OpenAI is seeking a valuation of $150 billion. According to sources familiar with the matter, the financing will be in the form of convertible notes. The high valuation is contingent on OpenAI changing its corporate structure and removing the earnings restriction for investors.
These developments mark a significant change for OpenAI, which was founded in 2015 as a nonprofit research project. The potential restructuring could make it easier for the company to attract investors and fund its costly artificial intelligence research. However, it also raises questions about OpenAI’s future direction and governance, particularly with regard to the company’s original nonprofit mission.
Apple Intelligence beta coming in October
Apple will release the beta version of Apple Intelligence with iOS 18.1 in October. The first features will be available in US English only, the company announced. Other English versions will follow in December for users in Australia, Canada and the UK. Chinese, French, Japanese and Spanish are planned for 2025.
Apple Intelligence is offered free of charge to users of the iPhone 15 Pro, iPhone 16 and iPhone 16 Pro, as well as Mac and iPad owners with an M1 chip or later. According to 9to5Mac’s Benjamin Mayo, features like notification summaries and photo editing in the Photos app will be rolled out first. Others will follow later.
B A C K G R O U N D
Investment in AI reaches dizzying heights
Investment in AI has reached unprecedented levels. As Nate Rattner and Tom Dotan report in their article for the Wall Street Journal , tech giants and investors are spending hundreds of billions of US dollars on AI. In the last quarter alone, Amazon, Microsoft, Alphabet, and Meta collectively invested more than $50 billion in capital expenditures, mostly on AI hardware.
The AI boom is also evident in other areas. Venture capital investment in AI startups has reached $64.1 billion so far this year. The number of data centers is growing rapidly, as is the energy demand for AI applications. As a major supplier of AI chips, Nvidia is benefiting enormously from this development. At the same time, the demand for AI specialists is rising sharply, while job openings in the technology sector as a whole are declining slightly.
The brute force phase of AI
Artificial intelligence that requires specialized hardware is doomed to fail, according to Gartner analyst Erick Brethenoux. In an interview with The Register , he explained that GPUs could also be affected. Brethenoux sees the current phase of AI as a “brute force” approach in which immature programming techniques are compensated for by powerful hardware.
He predicts that generative AI will be suitable for only about five percent of use cases, while 90 percent of attention is focused on it. Many companies have already realized that established AI techniques are often more efficient and cost-effective.
Instead, Brethenoux recommends “composite AI,” which combines generative AI with proven methods such as machine learning.
How trustworthy is Chatbot Arena?
Chatbot Arena, a benchmarking tool for AI models, has become very popular in the tech industry. As Kyle Wiggers reports on TechCrunch, companies like OpenAI and Google use the platform to test the performance of their chatbots. Millions of people visited the LMSYS website last year.
However, experts are critical of the validity of the benchmark. According to Yuchen Lin of the Allen Institute for AI, there is a lack of transparency regarding the skills tested. Also, the composition of the users rating the chatbots may not be representative. Mike Cook of Queen Mary University of London points out that Chatbot Arena provides relative ratings rather than empirical tests.
Despite these limitations, experts see the platform as a useful tool for gaining insight into the performance of AI models.
Overview of the potential and techniques of AI agents
This article by Michael Trestman at VentureBeat offers a comprehensive insight into the development of “Agentic AI”, the hoped-for next level of artificial intelligence. This new form of AI promises to act and make decisions autonomously, which could fundamentally change business processes. Trestman explains the basics of this technology and its potential impact on various industries.
The article covers several key aspects: It defines Agentic AI and its characteristics, describes potential application areas such as sales, marketing, cybersecurity, and IT operations, and discusses implementation challenges. It also presents specific examples of AI agents and their potential applications.
Trestman emphasizes that while this technology is still in its infancy, it has enormous potential to make businesses more efficient and agile.
AI pioneer Fei-Fei Li fascinated by “spatial intelligence”
Fei-Fei Li, known as the “godmother of AI,” has founded a startup called World Labs. The company aims to develop AI systems that can construct complete worlds with the physics, logic, and details of our physical reality. Despite an alleged slowdown in the AI sector, investors have already poured $230 million into the project, valuing it at $1 billion. This is reported by Steven Levy in an article for WIRED.
Li, who made a breakthrough in AI a decade ago with ImageNet, sees spatial intelligence as the next stage of development. World Labs plans to have a product on the market by 2025 that could be used by everyone from game developers to movie studios. High-profile investors such as Andreessen Horowitz and Nvidia are backing the project.
G L O S S A R Y
Chain of Thought
Chain of Thought is a concept in artificial intelligence that describes the ability of AI systems to solve complex problems step-by-step, much like humans do. This method allows AI models to explain their thought processes in a way that humans can understand.
Instead of just providing a final answer, the AI shows the individual steps of reasoning that led to that solution. This is particularly useful for tasks that require logical thinking, multi-step calculations, or complex reasoning.
By using Chain of Thought, AI systems can not only deliver more accurate results, but also make their decision-making more transparent. This makes it easier for users to understand the logic behind the AI's answers and identify potential errors.
In practice, Chain of Thought is often used in speech models and other generative AI systems to improve their problem-solving capabilities and make human-machine interaction more understandable.