In the news category, we keep you updated on the latest developments and trends within the AI and marketing world.

Here we share up-to-date information on important events, events, innovative developments, market shifts and more.

We go in depth and give you insights so that you always stay well informed and can make strategic decisions based on the latest information. After all, within marketing it is extremely important to keep up to date because developments are happening so fast.

Follow us and stay up to date.

, ,

Why the importance of AI policy and training is increasing

In recent years, the influence of artificial intelligence (AI) in the workplace has grown exponentially. According to the 2024 Annual Work Trend Index, 75% of employees now use AI at work. Interestingly, 46% of these users have only started using AI tools within the past six months. This means that a large proportion of workers had to get used to AI in a short period of time, without much experience or guidance.

Despite this rapid adoption of AI in business, the 2024 State of Marketing AI Report reveals that 75% of companies do not yet offer internal AI-focused education or training. Less than 40% of organizations have a clear AI policy in place. This means many employees are using AI without actually really knowing how to use it properly, which can lead to unknowingly sharing trusted information and not using the tools correctly.

So why is an AI policy so important?

Using AI without clear policies carries risks. Without agreements or guidelines, employees may unknowingly deploy tools in ways that do not align with business strategy or may even harm the company. AI, when applied correctly, is a powerful tool to increase efficiency and drive innovation. But without a proper plan or clear agreements, its use can have just the opposite effect.

An AI policy provides structure and clarity within the organization. It helps teams better understand which tools may or may not be used and how to use these tools effectively and responsibly. Such a policy can include agreements about:

  • Which AI tools may be used.
  • How data is processed and secured.
  • What ethical guidelines apply when using AI.

Developing an AI policy doesn't have to be complicated. We have a free template you can use. By adapting it to your organization's specific needs, you can quickly create a good policy that will help integrate AI responsibly into your daily processes.

What AI tools are commonly used?

An important part of an effective AI policy is understanding what AI tools are being used within your organization and what agreements have been made about them. Many companies do not have a clear overview of the tools their employees are using. This lack of overview can lead to unnecessary risks.

According to the recently released third edition of the Top 100 Gen AI Consumer Apps, the market for AI tools is still going strong. Nearly 30% of the companies listed are new since the previous report in March 2024. This rapid change demonstrates how dynamic the market is and shows how necessary it is for companies to constantly keep abreast of new developments.

A useful starting point for organizations is to look at the most widely used AI tools. The Top 50 AI Tools list provides a good overview of the tools that are currently popular, both for Web users and mobile apps. By reviewing this list, you can better determine which tools are right for your team and what training is needed to deploy them effectively.

A future with AI requires a good AI policy

AI offers unprecedented opportunities for businesses, but without proper guidelines, its use can quickly spiral out of control. Creating an AI policy and providing targeted training to employees are essential steps to deploying AI in a safe and responsible manner. By streamlining and guiding the use of AI tools within your team, you will ensure that your organization takes full advantage of the benefits of this technology without losing sight of the risks.

Consider establishing an AI policy and offering appropriate training. That way you'll be ready for the future, when AI will play an even bigger role in the workplace.

Take a leap forward in your marketing AI transformation every week

Every Friday, we bring you the latest insights, news and real-world examples on the impact of AI in the marketing world. Whether you want to improve your marketing efficiency, increase customer engagement, sharpen your marketing strategy or digitally transform your business. 'Marketing AI Friday' is your weekly guide.

Sign up for Marketing AI Friday for free.

Marketing AI Friday
/by
, ,

OpenAI back at #1 with update GPT-4o

In August 2024, OpenAI once again gained attention by launching the latest update to its GPT-4o model. This update, released on Aug. 8 brought the company back to the top of the AI rankings, tested by Chatbot Arena, after being briefly overtaken by Gemini 1.5.

What makes this return special is that OpenAI did not introduce an entirely new model, but rather thoroughly improved the existing GPT-4o model.

ChatGPT-4o is now ranked No. 1 in the following categories:

  • General
  • Mathematics
  • Coding
  • Complex prompts
  • Instructional follow-up
  • Longer questions
  • Multi-step interviews

Improved performance and better problem solving

The focus of the GPT-4o update was primarily on optimizing performance and addressing existing bugs.

Instead of choosing to launch an entirely new model, OpenAI chose to build on the strong foundation of GPT-4o and refine it further. They did this by thoroughly purifying the dataset, replacing outdated or inaccurate information with current and accurate data. As a result, the model has now taken a significant leap forward, both in speed and efficiency.

Complex prompts are answered better

One notable improvement of the updated GPT-4o model is how well it now handles complex prompts. While previous versions sometimes had trouble understanding complex prompts, with this update it seems to be very accurate and fast at handling complex issues. This makes the model especially useful for professionals and companies that use AI for tasks that require a high level of precision. These include analyzing complex data sets or developing sophisticated software.

Better instructional follow-up

Furthermore, GPT-4o is now also better able to accurately follow long and detailed instructions. This was an issue where previous versions of the model sometimes fell short, resulting in output errors. The update has addressed these shortcomings, so users can now rely on an AI that executes instructions more efficiently and consistently. This is especially valuable for companies using AI to automate day-to-day tasks and complex workflows.

Growing impact of GPT-4o

The improvements in GPT-4o not only strengthen OpenAI's position, but also provide tangible benefits to users worldwide. The model is now even more capable of performing a wide range of tasks with a higher degree of reliability and accuracy.

Signal to competitors

This update demonstrates the importance of continuous innovation within AI. Although Google's Gemini 1.5 temporarily took the lead, OpenAI's rapid response shows that continuous improvement of existing models is just as important as developing new ones.

Invitation to monthly AI inspiration sessions

Together with you, we would like to take a look at the most important AI developments, discuss the latest news and updates, share experiences, inspire each other, take advantage of what is possible and prepare ourselves as well as possible for the changes that are coming.

Would you like to attend? Then sign up.

Banner Online AI inspiration session
/by
,

Google's core algorithm update August 2024: What you need to know

When Google makes a major update in their algorithm, it has a major impact on how websites are ranked in Google and what position you come up with keywords. Google regularly adjusts the algorithm to ensure that users get the most relevant search results.

The August 2024 core algorithm update is one such update, in this article we share some findings about the update and advice on what to focus on.

Forums and User Generated Content (UGC) continue to do well

During the update now underway in August 2024, we see that forums and user-written content (UGC) still perform well in search results.

While this is not a new trend and has been around for some time, it is remarkable that this type of content continues to have so much value. Reddit is a good example of this. A small increase in visibility can quickly lead to many additional visitors for Reddit, as the platform already has a large number of users.

Niche and local businesses get more visibility

Another important change is that both niche and local businesses seem to be doing better in search results. This presents opportunities especially for SMEs and niche companies. Where previously large national players or multinationals were often at the top (partly due to a combination of amount of content, domain authority, through links and branding) we now see that smaller and medium-sized companies are also scoring better in national searches. This is a positive development as it also gives smaller companies a chance in the playing field.

Relevance of your website

Domain relevance also plays a larger role. It remains important that your domain name and the services you offer are consistent with other content on your website. Sites that get this right seem to hold their position(s) well or rank better in search results than before.

So while we see smaller and medium-sized companies doing well, that does not mean that all big brands are/will do worse. Brands that already have an established name and receive constant flow of authoritative backlinks continue to perform well in search results. These brands continue to do well because of their broad visibility both on and off the Internet and because of their a multi-channel approach.

Our findings

With an algorithm update, it is important not to react emotionally too quickly, whether you are positive or negative about the changes in terms of rankings. This is because it may take some time before all the effects are fully visible.

Recently, there have been many complaints about Google's performance, as some people feel the search results have become less relevant. Google seems to be taking these concerns seriously, especially given the emergence of new competitors such as SearchGPT, which means that in addition to SEO, AIO is going to become more important.

The August 2024 core update shows how important it is to create content that fits well with what your visitors are looking for. The success of forums and User Generated Content shows that people need valuable and real information. Therefore, it is important that your content is not only informative, but that it really responds to your visitor's needs.

What can you do?

Make sure that the content on your website is relevant to your services and matches your visitors' questions and expectations. You can do this, for example, by creating a content strategy, in which you create content that connects to different stages of the fan journey. It is also important that your content is well structured and easy to read, so that visitors can quickly find what they are looking for.

Google values websites that put their visitors first and provide the best user experience. By focusing on the quality of your content, you increase your chances of ranking well in search results regardless of future Google updates. Visitors will stay on your site longer and actually be interested in what you have to say, which can further improve your ranking.

Invitation to monthly AI inspiration sessions

Together with you, we would like to take a look at the most important AI developments, discuss the latest news and updates, share experiences, inspire each other, take advantage of what is possible and prepare ourselves as well as possible for the changes that are coming.

Would you like to attend? Then sign up.

Banner Online AI inspiration session
/by
, ,

Elon Musk's xAI launches Grok 2, an AI 'wild west'

Elon Musk's xAI has announced Grok 2, a new version of their AI chatbot available exclusively to X Premium and X Premium+ (formerly Twitter) subscribers. This introduction is startling and creates a "wild west" in the AI landscape.

We tested it and asked the AI itself to tell a bit about this introduction and we were positively surprised. In this article, we talk about Grok 2 and the important features of this AI model.

What is Grok 2?

Grok 2 is the latest chatbot from xAI, the successor to earlier versions within the Grok series. Released in beta in August 2024, this version offers many more features than its predecessors. Grok 2 can not only respond and reason smarter, but can now create images thanks to its collaboration with Black Forest Labs' FLUX.1 model.

Key features of Grok 2

Improved reasoning capabilities

Grok 2 is considered "better" compared to leading competitors, such as Meta and Anthropic models. According to benchmarks on LMSYS, Grok 2 performs particularly well, making it an attractive option for users looking for advanced AI features.

Image Generation

A notable new feature of Grok 2 is the ability to generate photorealistic images. This is made possible by its integration with FLUX.1. It can create largely uncensored images that can be both innovative and controversial. Users of X can now create images, which was not possible before.

Accessibility

Grok 2 and its smaller variant, Grok-2 mini, are only available to X Premium and Premium+ subscribers. This means that Grok 2's special features are only accessible to paying users, which creates a certain form of exclusivity within the user base.

Comparison with competitors

Although Grok 2 receives much praise for its performance, it has not yet broken the barrier of GPT-4. Nevertheless, it does belong to the class of "GPT-4 class" AI models, which means it can compete with the best AI models currently available.

Cultural impact

Especially because it can create controversial images, it has raised many questions about responsibility and ethics of AI. Since this technology can be used to create fake images of famous people, it creates discussion about how we use technology and what responsibility platforms have about what people do with their tools.

Grok 2 in action

To test whether Grok 2 can provide current and accurate information on topics other than itself, I asked where footballer Mbappé is currently playing and how his debut went. This was Wednesday night, so very topical. Grok 2 gave a correct answer, although the Dutch sometimes sounded a bit unnatural.

After testing several questions, the topicality and "playful" way of answering stand out in particular. In his responses, Grok refers exclusively to related posts on X, as in this example to Voetbal International. Smaller Twitter accounts are also mentioned, sometimes with only a few hundred followers.

Potential hazards and opportunities

Clearly, Grok 2 is trained on the content of X users, which ensures that it can provide up-to-date responses. However, this also carries risks, especially if the information is inaccurate. This development is especially interesting for companies that share newsworthy information and have an audience that is active on X.

Does this apply to your company? Then it might be worth getting active (again) on X and making sure your content is available for Grok's training data. That way, your posts can be included in Grok 2's results, helping you reach a wider audience.

Grok 2 vs. other chatbots

Compared to other chatbots such as ChatGPT, Grok 2 still seems to have its limitations, including its slower response time. Nevertheless, it is certainly an interesting development to keep an eye on and experiment further with. Especially since Elon Musk has already announced that Grok 3 will be launched by the end of this year. His ambition is that Grok 3 will be the most advanced language model, which will be at or above the level of the not yet launched GPT-5.

Trying out Grok 2

If you want to try Grok 2 for yourself, you can click on X Premium in your X settings and activate the Premium subscription there. Then through settings -> X Premium -> Grok, you can try out the Grok 2 mini (beta) version. Grok 2 is also available in the Netherlands from the beginning of this month.

Read more about Grok 2 here.

Invitation to monthly AI inspiration sessions

Together with you, we would like to take a look at the most important AI developments, discuss the latest news and updates, share experiences, inspire each other, take advantage of what is possible and prepare ourselves as well as possible for the changes that are coming.

Would you like to attend? Then sign up.

Banner Online AI inspiration session
/by
, ,

Google introduces Pixel 9 and Gemini Live

Early this week was the day: Google's much-anticipated Made by Google event. This event revolved entirely around artificial intelligence (AI). In the spotlight were the brand new Pixel 9 series and Gemini Live, Google's advanced AI assistant. With these announcements, Google is taking a clear step toward fully integrating AI into all their products, from smartphones to smart home devices. But what exactly does this mean for users? We discuss it in this article.

What's coming?

Gemini Live: The New Standard in AI Assistants

Gemini Live is more than a simple upgrade to Google Assistant. This new AI assistant is going to replace Google Assistant on Android phones. The idea behind Gemini Live is that you can have conversations naturally with your phone. You can ask questions, get help with various tasks and even pause in a conversation to pick up later where you left off. This makes Gemini Live a powerful tool for the people who want to get a lot out of their smartphone.

Gemini on Pixel: Contextual help at your fingertips

One notable feature of Gemini on the new Pixel phones is the ability to get contextual help simply by holding down the power button. Whether you're busy writing an email, planning your week, or learning a new language, you can let Gemini support you. This makes using your smartphone not only easier, but also much more efficient. No more struggling to find the right information, Gemini already has the answer ready for you.

Gemini Extensions: Multitasking gets easier than ever

Gemini Extensions makes AI even more accessible by integrating the assistant with various Google apps, such as Gmail, Maps, YouTube Music and Drive. This means you can multitask more efficiently without switching apps. Imagine this: while planning a route in Maps, you can simultaneously organize your emails and listen to your favorite music, all with Gemini's help. This integration makes the daily use of your devices a lot more productive.

Pixel Studio: Get creative anytime, anywhere

In addition to Gemini Live, Google is also introducing Pixel Studio, an app that uses AI to generate images. Whether you work on your phone or in the cloud, Pixel Studio allows you to create visual content with ease. This tool is ideal for anyone who wants to create attractive visuals quickly and easily without requiring in-depth knowledge of graphic design.

Call Notes and Reimagine: AI for efficiency and creativity

Call Notes and Reimagine are two other features that show how AI can improve our daily lives. Call Notes lets you summarize phone calls and create transcripts right on your phone. This not only ensures optimal privacy, but also a more efficient workflow, especially if you spend a lot of time on your phone.

Reimagine goes a step further by allowing you to edit parts of images with AI-generated content based on text. So there are many new opportunities for creative editing, without the need for complex software.

Magic Editor updates: Automatic suggestions for perfect photos

Another impressive feature Google is introducing is the Magic Editor. This tool offers automatic suggestions for cropping and adjusting photos. Whether you are a professional photographer or just want to enhance your vacation photos, Magic Editor ensures that your photos always look their best. This makes photo editing easier and faster.


With Gemini, Google is not just targeting phones. The new Nest Thermostat uses Gemini for more natural voice interaction, and Google TV Streamer will soon get Gemini features as well.

Gemini Live is first available to Gemini Advanced subscribers with an Android phone (in English). iOS and more languages will follow soon.

You can download and use the Gemini app on Android devices or you can activate Gemini in your Google app on an iPhone (via Settings -> Other Settings -> check 'Show Gemini'). Then in Home at the top next to 'Search' click on the blue Gemini icon. With that, you switch to Gemini.

IMG_7129

Invitation to monthly AI inspiration sessions

Together with you, we would like to take a look at the most important AI developments, discuss the latest news and updates, share experiences, inspire each other, take advantage of what is possible and prepare ourselves as well as possible for the changes that are coming.

Would you like to attend? Then sign up.

Banner Online AI inspiration session
/by
, ,

Is newcomer Flux an alternative to Midjourney or DALL-E?

We are getting more and more opportunities to create images with AI. Earlier this month we added Flux, developed by Black Forest Labs that raised $31 million during an investment round.

What is Flux and how does it work?

Flux focuses on speed and efficiency, generating images in real-time. This allows users to experiment immediately with different styles and concepts. The speed at which Flux works is one of the key features that sets it apart from other AI photo generators on the market. Users can create a visual within seconds, which is a huge advantage in creative processes where speed can be an important factor.

Control and flexibility with Flux

Flux also offers a wide range of options for control and flexibility in image editing. The interface is user-friendly and accessible, making it an attractive tool for both beginners and advanced users. In addition, it is possible to further edit and customize the generated images, which is a big plus for those still eager to make adjustments to their design.

Flux versus Midjourney and DALL-E: A comparison

Flux competes with established players such as Midjourney and DALL-E and claims it is already better in the following areas:

Visual quality

The sharpness, level of detail, color saturation and overall aesthetics of the generated images. Flux uses advanced algorithms that ensure higher precision and a more realistic end result.

Prompt-tracking

Flux excels in how well the AI is able to interpret instructions and descriptions (prompts) given by the user and translate them into an image. This means users get exactly what they envision.

Variation in size and proportion

Users can very easily generate images in different sizes and formats (e.g., square, rectangular, panorama), which is useful for various applications such as social media, websites or print.

Text in pictures

Flux stands out in particular for its ability to properly incorporate text into images. Where other AIs struggle to integrate readable and aesthetically pleasing text, Flux succeeds.

Diversity in output

Flux is capable of generating different types of images, varying in style, subject matter and complexity. This makes it a versatile tool for both artistic and commercial applications.

The different versions of Flux

Flux offers three different versions, each with its own unique advantages:

  • Pro: The top model that offers the best results in terms of image quality and functionality. Ideal for professionals looking for the best of the best.
  • Dev: The mid-range device that offers a balance between speed and quality. Suitable for users who value both speed and good results.
  • Schnell: The fastest model, but with the lowest quality. This is a good option if you need something simple quickly and do not necessarily require the highest quality.

ELO score: How does Flux perform in comparison?

Below you can see their ELO score relative to the other players. In the context of AI image generators, the ELO score is used to compare the quality and performance of different models. A higher ELO score means that the model performs better on a set of criteria.

First impressions of Flux: what do users think?

Flux still has plenty to prove, but first impressions are that the quality and ease of creating images are very good. See below an example of an image that was created within seconds.

Take a leap forward in your marketing AI transformation every week

Every Friday, we bring you the latest insights, news and real-world examples on the impact of AI in the marketing world. Whether you want to improve your marketing efficiency, increase customer engagement, sharpen your marketing strategy or digitally transform your business,'Marketing AI Friday' is your weekly guide.

Sign up for Marketing AI Friday for free.

Marketing AI Friday

/by
, ,

Gemini Pro 1.5: Google's best chatbot yet

Version naming is becoming more and more particular, but in relative silence, Google recently released the experimental version of Gemini Pro 1.5. This new version builds on the earlier Gemini 1.5 Pro and introduces several improvements that both users and developers can appreciate.

What's new in Gemini Pro 1.5?

Gemini Pro 1.5 offers a number of noteworthy improvements over its predecessor. Below is an overview of the main innovations:

Improved factual accuracy

One of the most significant updates is the increased accuracy of the answers generated by the model. This is because the underlying datasets have been updated with more recent information, allowing the model to provide more reliable and more informed answers. This is particularly important for users who rely on current and accurate information, such as when making business decisions or academic research.

Enhanced reasoning

Gemini Pro 1.5 has also received an upgrade in the area of reasoning. This means that the model is now better able to understand and respond appropriately to complex questions and scenarios. Where previous versions sometimes struggled to interpret complex question sets, this version shows more consistency in delivering logical and correct answers.

Enhanced coding

For developers, this is an important update: the model's ability to generate code in various programming languages has been greatly improved. This makes Gemini Pro 1.5 more useful not only for simple tasks such as writing basic scripts, but also for generating code in various programming languages.

Improved instruction tracking

Another important aspect of this version is the model's improved ability to follow instructions and constraints. This means that users have more control over the model's output and the results more accurately match their specific requirements.

Gemini Pro 1.5: the new frontrunner

In addition to the technology improvements, Google has reached the top of the Chatbot Arena Leaderboard for the first time with Gemini Pro 1.5. This platform allows chatbot developers to have their creations compete in various categories, with user experience at its core. The leaderboard provides a ranking of the best chatbots based on performance in various domains.

With this version of Gemini, Google manages to capture the leading position, indicating that the company is making serious strides in the world of AI-driven chatbots. In terms of multilingualism and technical skills such as math, complex prompts and coding, Gemini Pro 1.5 scores particularly high.

Screenshot 2024-08-08 at 15.39.39

Key positions on the leaderboard are as follows:

  • Overall: #1
  • Mathematics: #1-3
  • Instructional tracking: #1-2
  • Coding: #3-5
  • Complex prompts (English): #2-5

These results reflect the versatility and power of Gemini Pro 1.5. Especially in demanding tasks where in-depth knowledge and precise execution are required.

Test it yourself

If you have access to Google AI Studio, you can try this version of Gemini Pro 1.5 and experience the improvements for yourself.

Take a leap forward in your marketing AI transformation every week

Every Friday, we bring you the latest insights, news and real-world examples on the impact of AI in the marketing world. Whether you want to improve your marketing efficiency, increase customer engagement, sharpen your marketing strategy or digitally transform your business,'Marketing AI Friday' is your weekly guide.

Sign up for Marketing AI Friday for free.

Marketing AI Friday
/by
, ,

Will OpenAI soon launch the forward-thinking 'Strawberry' project or will it lose momentum?

In recent months, OpenAI announced several impressive updates such as Sora, a new Voice Mode and SearchGPT, but nothing has really been delivered yet. There are increasing critical voices about this. Only Voice Mode has been made available to a limited group of real users. The first sounds about Voice Mode are very positive and several experts even wonder if this feature is not already using the successor of ChatGPT-4o.

Now there are also rumors of the launch of a project codenamed "Strawberry" (formerly Q-Star).

Top executive Sam Altman added to this a few days ago with this tweet.

IMG_6967

Although nothing has been officially confirmed and details are kept strictly secret within OpenAI, there are several rumors and speculations circulating about "Strawberry. 

What is Strawberry from OpenAI?

Strawberry is said to be a new AI technology that OpenAI has been working on for some time, with which the company intends to push the boundaries of AI. It goes far beyond the current capabilities of ChatGPT-4 and DALL-E and is seen as an important step toward Artificial General Intelligence (AGI). This is the point where AI systems not only become more intelligent than humans, but can also perform most economically valuable tasks better. 

What can we expect?

A revolution in AI creativity: Strawberry is said to be pushing the boundaries of creative AI, allowing users to generate even more complex and realistic content.

Augmented intelligence: It offers much more complex and nuanced form of AI intelligence and is able to plan, reason, learn and understand in a way we do not know today. It could be a big step forward AGI.

Independent Internet use: It can independently search the Internet and gather information. Just as a human would, but much faster and more efficiently.

Long-Term Tasks: It can perform complex tasks and projects that take longer periods of time and require continuous adjustments and planning. 

This development also raises important ethical questions. As AI models become more autonomous and powerful, it is crucial to adopt ethical frameworks and responsible development methods. This is essential to mitigate potential risks and unintended consequences. OpenAI faces the challenge of striking a balance between encouraging innovation and ensuring that these advances are used responsibly and ethically for the benefit of society. 

Recently, OpenAI has also been remarkably communicating more about their approach around secure AI. 

Where did the name "Strawberry" come from? 

It has not been disclosed where the name "Strawberry" comes from, but it seems to be a name that was not chosen lightly and can possibly be traced back to Elon Musk. Musk now competes with OpenAI through his own AI model Grok, but he was also one of the 11 founders of OpenAI at the time, with which they particularly wanted to compete with Deepmind (where Musk was also previously one of the early investors, by the way, before it was acquired by Google in 2014). 

While working with Sam Altman at OpenAI before he left the organization in 2018, Elon Musk used "strawberries" in an anecdote to emphasize the need for safety and ethical considerations in AI development. He reportedly used the example of an AI system tasked with becoming the best at making strawberries.

The system might then decide that the best way to do this is to produce as many strawberries as possible, at any cost. In an extreme scenario, the AI system would eventually devote so many resources to producing strawberries that it could turn the entire planet into a giant strawberry plantation, which of course would be catastrophic for human civilization and the ecosystem. 

This example was used by Musk to demonstrate how an AI system, without proper constraints and ethical guidelines, could adopt a very narrow and destructive interpretation of a seemingly innocuous task. It illustrates the point that AI, if not properly controlled or guided, could pursue goals in a way that is ultimately harmful to humanity. Elon Musk regularly makes negative comments about OpenAI and a few days ago reopened the lawsuit he had previously filed against OpenAI and Sam Altman. The name "Strawberry" may have stemmed from that anecdote at the time and be a form of reaction to Musk.

Also hefty challenges for OpenAI

OpenAI also seems to face quite a few challenges.

For example, co-founder and President Greg Brockman announced this week that he is taking a sabbatical until the end of this year. On the same day, co-founder John Schulman announced that he was leaving OpenAI for a role at competitor Anthropic (Claude), where several OpenAI people have previously gone. Peter Deng, a key product leader, also appears to be gone. Meanwhile, of the 11 founders at the time, only three remain. This raises questions about internal support for the company's current direction.

In addition, OpenAI appears headed for a $5 billion loss this year, major shareholder Microsoft (49% stake) has added OpenAI to its list of direct competitors, and the company must continue to invest billions to train new language models on larger and higher-quality datasets. This means they also need many more GPUs (Graphics Processing Unit), special chips designed to process graphics tasks quickly and efficiently,

For comparison, here are some figures on how many GPUs different companies are using for their AI models:

GPT-4 is trained on 25,000 GPUs (OpenAI)
Llama3 - 16,000 GPUs (Meta)
Grok 2 - 20,000 GPUs (Elon Musk)
Grok 3 - 100,000 GPUs (Elon Musk)
Llama4 -160,000 GPUs. (Meta)

This comparison shows how intensive the training process is and how many resources are needed to develop advanced AI models.

Other challenges

An additional challenge for OpenAI is that new and more extensive data is much more limited access for them compared to Google (Chrome web browser, YouTube, Gmail, Search, Drive etc.) and Meta (Facebook, Messenger, Instagram, Whatsapp and Threads in particular).

This situation does not only apply to large technology companies. Even for smaller organizations and individual developers, the quality and quantity of available data, along with time and resources for training and development, determine the effectiveness of AI applications. 

It will be interesting months to see if OpenAI can maintain its lead and take big steps forward again. Or whether it will lose momentum and begin to lose its position as a forerunner.

Given these rapid developments, it is important to be flexible in your AI strategy. Using and developing AI applications around a single company can be risky.

A better approach is to develop your AI infrastructure in a modular fashion, so that you can easily replace or modify components AI tools and applications.

Take a leap forward in your marketing AI transformation every week

Every Friday, we bring you the latest insights, news and real-world examples on the impact of AI in the marketing world. Whether you want to improve your marketing efficiency, increase customer engagement, sharpen your marketing strategy or digitally transform your business,'Marketing AI Friday' is your weekly guide.

Sign up for Marketing AI Friday for free.

Marketing AI Friday
/by
,

Google Lets Users Decide About Cookies

The change in cookie policy

Google recently shared that they will not completely abolish third-party cookies after all as previously planned. Google still plans to abolish third-party cookies, but this will be done gradually through the Privacy Sandbox, which offers alternatives to tracking. Instead, users will have the choice of whether to allow cookies or not. This raises questions about the implications for advertisers and the protection of personal data. In this article, we talk about the impact on advertisers and users and the possible consequences for in the future.

Apple's Example and Google's Approach

In 2021, Apple introduced the App-Tracking Transparency (ATT) feature, allowing users to choose whether they wanted to be tracked by apps. This made it a lot harder for advertisers to track users. In fact, the opportunities to do so became a lot less. This had a big financial impact on Meta. Google offers privacy-friendly alternatives to third-party cookies through Privacy Sandbox, but does not explicitly ask users for permission for every tracking action the way Apple's ATT does.

The impact on users & advertisers

Consumer preferences and privacy

An important factor in this change is consumer attitudes toward privacy. Indeed, users have become much more aware and caring about this. In fact, data shows that a large number of users are blocking cookie tracking. According to online sources, about 70% of consumers are currently not being tracked by cookies. This is due to blocking by the user themselves or by using browsers such as Apple's Safari, which generally blocks cookies by default.

Advantages and disadvantages for advertisers

If cookies disappear or if many users choose to block them, this could reduce the effectiveness of online ads and reduce advertisers' revenues. Advertisers will therefore have to be more transparent about their data usage and may have to invest in server-side tracking, among other things.

Google's Privacy Sandbox is a new initiative to help advertisers still show targeted ads while protecting users' privacy. Reactions to this system have been mixed. Some worry about how well it will work and the potential impact on privacy.

What does this mean for the future?

Google's decision to let users make their own choice about cookies indicates the importance of privacy. This means advertisers will adapt to this situation, especially when the exact future of online tracking is uncertain and in which there will be more emphasis on transparency and privacy. This shift may affect that way we advertise, with more focus on first-party data.

While first-party cookies are essential for the basic functionality and user experience of a website, third-party cookies are often used for targeted advertising and tracking. If you want to learn more about the differences between first-party and third-party cookies and how they affect privacy and advertising, you can read our earlier article on first-party and third-party cookies. In this article, we discuss in detail the differences between the two types of cookies and their impact on.

Getting Started with Server Side Tracking

First party cookies and third party cookies have their own applications and challenges, but with increasing restrictions on third party cookies, a new approach is needed. Starting to apply server side tracking is a good option in this. Would you like to get help with this? Please contact us.

/by
, ,

Meta launches Llama 3.1: a big step forward with open source AI

Meta took a big step forward early this week. With the launch of Llama 3. 1, the company introduced an AI model that is as powerful as the most advanced systems such as ChatGPT-4o and Claude Sonnet 3.5, but freely accessible to all.

What is Llama 3.1?

Llama 3.1 is a large language model (LLM) developed by Meta. The main features of this model:

  1. Size: the model has three sizes: 405B, 70B and 8B, where B refers to the number of parameters. The 405B, for example, is 405 billion parameters and one of the largest models available. It can handle up to 128,000 tokens.
  2. Architecture: it is based on transformer architecture, similar to GPT models such as ChatGPT.
  3. Open source: unlike most other models, anyone can use Llama 3.1 for free and modify it to their liking. Also, you can install it on your own computer and also use it offline (e.g. via LM Studio). For this, the 8B version is best suited.
  4. Multimodal: Although it can currently process only text and files as input and output, other forms such as images, video and audio will be added in the future.

What makes it different?

The big difference from the other models is that Llama 3.1 is open source. The advantages of this are:

  1. Adaptability: researchers and developers can fine-tune the model for specific tasks.
  2. Transparency: the operation of the model can be thoroughly studied, which is important for ethical AI development, among other things.
  3. Accessibility: everyone will have access to advanced AI technology.
  4. Innovation: the wide availability and ability to build on it can lead to many new applications. 

However, open source AI models also have challenges, such as misuse and the need for careful regulation. After all, anyone can develop products using this language model, including people with not so good intentions.

Llama 3.1 is not yet available via Meta AI in Europe, but you can test it via, for example, Groq, select it as a language model within Perplexity, or run it locally via LM Studio.

Below you can see how Llama 3.1 performs against other models such as ChatGPT4o and Claude 3.5 Sonnet.


Can it really compete with ChatGPT, Claude and Gemini?

It is to be expected that Meta will become a dominant player in the AI landscape. One of the big advantages they have is that they make money like water from their social media platforms and do not necessarily need to make money from AI. They can integrate the AI into their existing platforms such as Facebook, Instagram and Whatsapp, as well as in collaboration with partners such as the Ray-Ban smart glasses.

Also, open source is attractive for engaging many developers. In addition, their Chief AI Scientist is Yann LeCun, one of the leading AI researchers. You will also come across his name in this comprehensive overview of leading individuals and companies in the AI landscape.

Mark Zuckerberg is willing to invest billions of dollars to democratize the market for advanced AI models. With this strategy, he is also trying to undermine the core sales of competitors such as OpenAI and Anthropic and thwart the emerging market potential for Google, Microsoft and Amazon by giving away for free what they sell. 

The race to the next breakthrough

As Meta Llama 3.1 launches, other major breakthroughs are expected soon. OpenAI's GPT-4 is already 17 months old. In that time, other companies have invested billions to create similar models, but no one has definitively surpassed GPT-4. Will this Gemini 2, Claude 4, Llama 4 or perhaps Elon Musk's new Grok model (about which he himself announced in December that it is the most powerful AI model in the world) succeed?

Or will OpenAI make such a leap forward with the successor to ChatGPT-4 that they will once again set the standard for the time to come? And what level of their classification system will they be at? 

Time will tell, but we probably won't have to wait long for the answer. Almost immediately after the announcement of Llama 3.1, OpenAI already responded by announcing that their latest model GPT-4o mini is now also available for fine-tuning and that the first 2M tokens are free until Sept. 23.

In addition, OpenAI has made several announcements in recent months that have not yet been made available such as Sora, Voice Mode and other applications of ChatGPT-4o. All of these could be launched soon.

Regardless of which model leads the way (for a while), our advice remains to experiment with different AI tools and use the ones that best suit your specific tasks and needs.

Take a leap forward in your marketing AI transformation every week

Every Friday, we bring you the latest insights, news and real-world examples on the impact of AI in the marketing world. Whether you want to improve your marketing efficiency, increase customer engagement, sharpen your marketing strategy or digitally transform your business. 'Marketing AI Friday' is your weekly guide.

Sign up for Marketing AI Friday for free.

Marketing AI Friday
/by