July's Reality Check on AI Promises

July's Reality Check on AI Promises

July's Reality Check on AI Promises

July's Reality Check on AI Promises

Tips and Tricks

Tips and Tricks

8. aug. 2024

8. aug. 2024

Unfortunately for venture capitalists, there can’t be an AI breakthrough every month. Hereby, we round-up July – a month of AI research, criticism, and refinement. 

Per usual, we will dive into general advancements and research of artificial intelligence with a focus on relevant tools for marketers that help you make an informed decision about the usage of these tools and platforms. 

OpenAI developments, AI search race, humanoid tweets and what it has to do with marketers

July started with OpenAI’s suspension of ChatGPT's internet access due to user’s misuse.

OpenAI stops internet access

The main issue was that people used ChatGPT to access copyrighted content; for example, providing ChatGPT with a link to a pay-for-access article and asking it to print it. As a result, many people bypassed paid media. 

However, now it’s resolved. You still can access the internet, but not in the following instances:

  1. Sensitive or Private Information

  2. Copyrighted Material

  3. Malicious or Harmful Intent

  4. User's Specific Instructions

Talking about searches, OpenAI introduced a new prototype tool – SearchGPT. SearchGPT is an AI-powered search engine developed by OpenAI, designed to provide real-time access to online information. It aims to enhance the search experience by offering synthesized answers rather than just a list of links, incorporating features such as follow-up questions and visual answers.

Which sounds familiar… Perplexity AI, this you?

SearchGPT is Perplexity's copycat?

Although very similar, SearchGPT claims to be more visually pleasing and publisher friendly with the end goal of further integration with their flagship product, ChatGPT. 

For the marketers and business, big tech’s course on AI search assistants raises an important question:

What will marketing in the era of AI look like? 

A report from Ericsson indicated that 75% of consumers anticipate a future where marketing efforts will shift from targeting humans to focusing on AI assistants that will handle shopping decisions. This shift could fundamentally alter the advertising landscape.

And to think that already people are already so used to AI, they don’t distinguish it from human speech. Another scholarly study published from Science Advances this month found that “humans cannot distinguish between tweets generated by GPT-3 and written by real Twitter users.”

As marketers, we can’t help but wonder: what was that miraculous prompt that generated believable tweets and not some default AI-extravaganza?! 😂

Average marketer's chat with LLM

But jokes aside, this month’s studies call for a discussion on ethical production of media through the easily scalable ai-solutions.

Quick showreel of creative AI-tech news

Leonardo AI is acquired by Canva

Yes, we were shocked too! Canva has acquired Leonardo AI, a startup specializing in image AI generation. This acquisition aims to integrate Leonardo's capabilities into Canva's Magic Studio suite. With this integration, Canva users will benefit from enhanced tools for creating compelling visuals, making the design process more efficient and creative.

Stability AI rapidly advances

This month, Stability AI rolled out a couple of cool features. Firstly, we got a Stable Doodle, which is a fascinating tech turning your doodles and sketches into a generated image. Sounds pretty cool!

Moreover, they also announced Stable Video 4D, a model which turns video into a 4D image. Currently under the research, it already can generate 5 frames across 8 angles in 40 seconds! This is an insane breakthrough for VR, games development, and video production.

Wepost updates its platform!

Exciting news! This summer, Wepost has completely revamped our technology, transitioning from a prototype version to a fully self-serve tool. While we are still fine-tuning some parts of the process, we aim to launch our fully self-serve and freemium platform to the world in just 8 weeks.

New platform sneak peak

Be the first to try out the new platform when it goes live! Sign up for our waitlist today and stay ahead of the curve.

Discourse on the future of Generative AI

The end of July saw a significant debate on the capabilities of generative AI, sparked by an influential piece by Gary Marcus. This discussion highlighted some crucial aspects of AI development and its future trajectory.

Understanding AI: To fully grasp AI's potential, it's essential to understand how this technology works. Currently, we're in the first stage of generative AI, which is essentially machine learning. Large Language Models (LLMs) like ChatGPT don't possess human-like intelligence but are trained on vast amounts of data to predict the most likely word based on previous words.

Stages of AI Development: AI, AGI, ASI.

The weakness in the current stage is apparent through a simple test. Try asking your favorite LLM a modified version of any famous riddle. Here is an example: 

ChatGPT is hallucinating, again.

As you see, ChatGPT is not able to reason out of a problem. Instead it generates an answer that would resemble a typical answer to the riddle.  

Many experts, including Marcus, believe that achieving true intelligence (AGI) requires more than just feeding LLMs with data. He cites Google DeepMind's recent breakthrough as an example, where their AI achieved silver-medal standard in solving International Mathematical Olympiad problems by combining an intuitive LLM with a deliberative symbolic engine.

This approach aims to balance spontaneity with formal logic, combining both inductive and deductive thinking. It's reminiscent of Kahneman's System I (quick intuition) and System II (deliberative reasoning), integrating neural networks with explicit symbolic systems that use formal logic.

A picture depicting AlphaGeometry, combining Language Model and Symbolic Engine

Differing Perspectives: Not all experts agree with this hybrid approach. Some, like Erik Larson, argue that deductive reasoning can lead to incorrect conclusions, especially in a rapidly changing world. They emphasize the importance of understanding context, which companies like OpenAI are trying to achieve through deep learning.

While there's no consensus on the best path forward, the AI community is collectively working on advancing AI through various approaches. The ultimate goal is to create systems that can reason effectively in context, combining the strengths of different AI methodologies.

What does it all mean for the marketers?

Simply that you should know your tools – their current limitations and future potential.

The primary weakness of current generative AI is reliability. Marketers must rigorously fact-check AI-generated content to avoid spreading misinformation. Implement a verification process for all AI-produced materials before publication.

As AI-powered search engines like SearchGPT emerge, marketers need to adapt their SEO strategies. Focus on creating content that provides comprehensive, contextual information rather than just targeting specific keywords.

Stay informed about AI advancements and their potential impact on consumer behavior. Be prepared to adapt your marketing strategies as AI assistants potentially transform how people interact with media and make purchasing decisions.

Did you enjoy the read?
Then, subscribe to our newsletter to get the latest news and discussions about AI marketing and more!

Unfortunately for venture capitalists, there can’t be an AI breakthrough every month. Hereby, we round-up July – a month of AI research, criticism, and refinement. 

Per usual, we will dive into general advancements and research of artificial intelligence with a focus on relevant tools for marketers that help you make an informed decision about the usage of these tools and platforms. 

OpenAI developments, AI search race, humanoid tweets and what it has to do with marketers

July started with OpenAI’s suspension of ChatGPT's internet access due to user’s misuse.

OpenAI stops internet access

The main issue was that people used ChatGPT to access copyrighted content; for example, providing ChatGPT with a link to a pay-for-access article and asking it to print it. As a result, many people bypassed paid media. 

However, now it’s resolved. You still can access the internet, but not in the following instances:

  1. Sensitive or Private Information

  2. Copyrighted Material

  3. Malicious or Harmful Intent

  4. User's Specific Instructions

Talking about searches, OpenAI introduced a new prototype tool – SearchGPT. SearchGPT is an AI-powered search engine developed by OpenAI, designed to provide real-time access to online information. It aims to enhance the search experience by offering synthesized answers rather than just a list of links, incorporating features such as follow-up questions and visual answers.

Which sounds familiar… Perplexity AI, this you?

SearchGPT is Perplexity's copycat?

Although very similar, SearchGPT claims to be more visually pleasing and publisher friendly with the end goal of further integration with their flagship product, ChatGPT. 

For the marketers and business, big tech’s course on AI search assistants raises an important question:

What will marketing in the era of AI look like? 

A report from Ericsson indicated that 75% of consumers anticipate a future where marketing efforts will shift from targeting humans to focusing on AI assistants that will handle shopping decisions. This shift could fundamentally alter the advertising landscape.

And to think that already people are already so used to AI, they don’t distinguish it from human speech. Another scholarly study published from Science Advances this month found that “humans cannot distinguish between tweets generated by GPT-3 and written by real Twitter users.”

As marketers, we can’t help but wonder: what was that miraculous prompt that generated believable tweets and not some default AI-extravaganza?! 😂

Average marketer's chat with LLM

But jokes aside, this month’s studies call for a discussion on ethical production of media through the easily scalable ai-solutions.

Quick showreel of creative AI-tech news

Leonardo AI is acquired by Canva

Yes, we were shocked too! Canva has acquired Leonardo AI, a startup specializing in image AI generation. This acquisition aims to integrate Leonardo's capabilities into Canva's Magic Studio suite. With this integration, Canva users will benefit from enhanced tools for creating compelling visuals, making the design process more efficient and creative.

Stability AI rapidly advances

This month, Stability AI rolled out a couple of cool features. Firstly, we got a Stable Doodle, which is a fascinating tech turning your doodles and sketches into a generated image. Sounds pretty cool!

Moreover, they also announced Stable Video 4D, a model which turns video into a 4D image. Currently under the research, it already can generate 5 frames across 8 angles in 40 seconds! This is an insane breakthrough for VR, games development, and video production.

Wepost updates its platform!

Exciting news! This summer, Wepost has completely revamped our technology, transitioning from a prototype version to a fully self-serve tool. While we are still fine-tuning some parts of the process, we aim to launch our fully self-serve and freemium platform to the world in just 8 weeks.

New platform sneak peak

Be the first to try out the new platform when it goes live! Sign up for our waitlist today and stay ahead of the curve.

Discourse on the future of Generative AI

The end of July saw a significant debate on the capabilities of generative AI, sparked by an influential piece by Gary Marcus. This discussion highlighted some crucial aspects of AI development and its future trajectory.

Understanding AI: To fully grasp AI's potential, it's essential to understand how this technology works. Currently, we're in the first stage of generative AI, which is essentially machine learning. Large Language Models (LLMs) like ChatGPT don't possess human-like intelligence but are trained on vast amounts of data to predict the most likely word based on previous words.

Stages of AI Development: AI, AGI, ASI.

The weakness in the current stage is apparent through a simple test. Try asking your favorite LLM a modified version of any famous riddle. Here is an example: 

ChatGPT is hallucinating, again.

As you see, ChatGPT is not able to reason out of a problem. Instead it generates an answer that would resemble a typical answer to the riddle.  

Many experts, including Marcus, believe that achieving true intelligence (AGI) requires more than just feeding LLMs with data. He cites Google DeepMind's recent breakthrough as an example, where their AI achieved silver-medal standard in solving International Mathematical Olympiad problems by combining an intuitive LLM with a deliberative symbolic engine.

This approach aims to balance spontaneity with formal logic, combining both inductive and deductive thinking. It's reminiscent of Kahneman's System I (quick intuition) and System II (deliberative reasoning), integrating neural networks with explicit symbolic systems that use formal logic.

A picture depicting AlphaGeometry, combining Language Model and Symbolic Engine

Differing Perspectives: Not all experts agree with this hybrid approach. Some, like Erik Larson, argue that deductive reasoning can lead to incorrect conclusions, especially in a rapidly changing world. They emphasize the importance of understanding context, which companies like OpenAI are trying to achieve through deep learning.

While there's no consensus on the best path forward, the AI community is collectively working on advancing AI through various approaches. The ultimate goal is to create systems that can reason effectively in context, combining the strengths of different AI methodologies.

What does it all mean for the marketers?

Simply that you should know your tools – their current limitations and future potential.

The primary weakness of current generative AI is reliability. Marketers must rigorously fact-check AI-generated content to avoid spreading misinformation. Implement a verification process for all AI-produced materials before publication.

As AI-powered search engines like SearchGPT emerge, marketers need to adapt their SEO strategies. Focus on creating content that provides comprehensive, contextual information rather than just targeting specific keywords.

Stay informed about AI advancements and their potential impact on consumer behavior. Be prepared to adapt your marketing strategies as AI assistants potentially transform how people interact with media and make purchasing decisions.

Did you enjoy the read?
Then, subscribe to our newsletter to get the latest news and discussions about AI marketing and more!

Unfortunately for venture capitalists, there can’t be an AI breakthrough every month. Hereby, we round-up July – a month of AI research, criticism, and refinement. 

Per usual, we will dive into general advancements and research of artificial intelligence with a focus on relevant tools for marketers that help you make an informed decision about the usage of these tools and platforms. 

OpenAI developments, AI search race, humanoid tweets and what it has to do with marketers

July started with OpenAI’s suspension of ChatGPT's internet access due to user’s misuse.

OpenAI stops internet access

The main issue was that people used ChatGPT to access copyrighted content; for example, providing ChatGPT with a link to a pay-for-access article and asking it to print it. As a result, many people bypassed paid media. 

However, now it’s resolved. You still can access the internet, but not in the following instances:

  1. Sensitive or Private Information

  2. Copyrighted Material

  3. Malicious or Harmful Intent

  4. User's Specific Instructions

Talking about searches, OpenAI introduced a new prototype tool – SearchGPT. SearchGPT is an AI-powered search engine developed by OpenAI, designed to provide real-time access to online information. It aims to enhance the search experience by offering synthesized answers rather than just a list of links, incorporating features such as follow-up questions and visual answers.

Which sounds familiar… Perplexity AI, this you?

SearchGPT is Perplexity's copycat?

Although very similar, SearchGPT claims to be more visually pleasing and publisher friendly with the end goal of further integration with their flagship product, ChatGPT. 

For the marketers and business, big tech’s course on AI search assistants raises an important question:

What will marketing in the era of AI look like? 

A report from Ericsson indicated that 75% of consumers anticipate a future where marketing efforts will shift from targeting humans to focusing on AI assistants that will handle shopping decisions. This shift could fundamentally alter the advertising landscape.

And to think that already people are already so used to AI, they don’t distinguish it from human speech. Another scholarly study published from Science Advances this month found that “humans cannot distinguish between tweets generated by GPT-3 and written by real Twitter users.”

As marketers, we can’t help but wonder: what was that miraculous prompt that generated believable tweets and not some default AI-extravaganza?! 😂

Average marketer's chat with LLM

But jokes aside, this month’s studies call for a discussion on ethical production of media through the easily scalable ai-solutions.

Quick showreel of creative AI-tech news

Leonardo AI is acquired by Canva

Yes, we were shocked too! Canva has acquired Leonardo AI, a startup specializing in image AI generation. This acquisition aims to integrate Leonardo's capabilities into Canva's Magic Studio suite. With this integration, Canva users will benefit from enhanced tools for creating compelling visuals, making the design process more efficient and creative.

Stability AI rapidly advances

This month, Stability AI rolled out a couple of cool features. Firstly, we got a Stable Doodle, which is a fascinating tech turning your doodles and sketches into a generated image. Sounds pretty cool!

Moreover, they also announced Stable Video 4D, a model which turns video into a 4D image. Currently under the research, it already can generate 5 frames across 8 angles in 40 seconds! This is an insane breakthrough for VR, games development, and video production.

Wepost updates its platform!

Exciting news! This summer, Wepost has completely revamped our technology, transitioning from a prototype version to a fully self-serve tool. While we are still fine-tuning some parts of the process, we aim to launch our fully self-serve and freemium platform to the world in just 8 weeks.

New platform sneak peak

Be the first to try out the new platform when it goes live! Sign up for our waitlist today and stay ahead of the curve.

Discourse on the future of Generative AI

The end of July saw a significant debate on the capabilities of generative AI, sparked by an influential piece by Gary Marcus. This discussion highlighted some crucial aspects of AI development and its future trajectory.

Understanding AI: To fully grasp AI's potential, it's essential to understand how this technology works. Currently, we're in the first stage of generative AI, which is essentially machine learning. Large Language Models (LLMs) like ChatGPT don't possess human-like intelligence but are trained on vast amounts of data to predict the most likely word based on previous words.

Stages of AI Development: AI, AGI, ASI.

The weakness in the current stage is apparent through a simple test. Try asking your favorite LLM a modified version of any famous riddle. Here is an example: 

ChatGPT is hallucinating, again.

As you see, ChatGPT is not able to reason out of a problem. Instead it generates an answer that would resemble a typical answer to the riddle.  

Many experts, including Marcus, believe that achieving true intelligence (AGI) requires more than just feeding LLMs with data. He cites Google DeepMind's recent breakthrough as an example, where their AI achieved silver-medal standard in solving International Mathematical Olympiad problems by combining an intuitive LLM with a deliberative symbolic engine.

This approach aims to balance spontaneity with formal logic, combining both inductive and deductive thinking. It's reminiscent of Kahneman's System I (quick intuition) and System II (deliberative reasoning), integrating neural networks with explicit symbolic systems that use formal logic.

A picture depicting AlphaGeometry, combining Language Model and Symbolic Engine

Differing Perspectives: Not all experts agree with this hybrid approach. Some, like Erik Larson, argue that deductive reasoning can lead to incorrect conclusions, especially in a rapidly changing world. They emphasize the importance of understanding context, which companies like OpenAI are trying to achieve through deep learning.

While there's no consensus on the best path forward, the AI community is collectively working on advancing AI through various approaches. The ultimate goal is to create systems that can reason effectively in context, combining the strengths of different AI methodologies.

What does it all mean for the marketers?

Simply that you should know your tools – their current limitations and future potential.

The primary weakness of current generative AI is reliability. Marketers must rigorously fact-check AI-generated content to avoid spreading misinformation. Implement a verification process for all AI-produced materials before publication.

As AI-powered search engines like SearchGPT emerge, marketers need to adapt their SEO strategies. Focus on creating content that provides comprehensive, contextual information rather than just targeting specific keywords.

Stay informed about AI advancements and their potential impact on consumer behavior. Be prepared to adapt your marketing strategies as AI assistants potentially transform how people interact with media and make purchasing decisions.

Did you enjoy the read?
Then, subscribe to our newsletter to get the latest news and discussions about AI marketing and more!

Gjør historiene dine
til virkelighet.

Wepost © 2024.

Stensberggata 29, 0170 Oslo. Norway.

Org.number 930 324 787

Gjør historiene dine
til virkelighet.

Gjør historiene dine
til virkelighet.