In the dynamic landscape of AI marketing innovations, staying abreast of the latest advancements is critical for understanding how they shape our world. With countless sources of information, it's easy to feel daunted by the volume of updates. To simplify your discovery process, our team diligently sifts through resources, conducts meticulous research, and experiments with the newest tools. We share these findings in our weekly podcast, Artificially Intelligent Marketing, released every Friday.
In the most recent episode, we delve into a myriad of enthralling subjects, including:
- Anthropic's massive funding round: A glimpse into the future of AI assistants
- Windows 11 Copilot: Microsoft's AI personal assistant breakthrough
- Europe's AI pact with Google: An overview of the temporary agreement to mitigate AI risks
- Generative AI Beta in Photoshop: The new, accessible tool for creative AI generation
- Google's generative AI integration into ads and image assets: A transformative addition to advertising
- Tool of the week: Exploring the synergies between Zapier and OpenAI
You can listen to episode 12 below, or subscribe on your favourite podcast platform. If you're more of a reader, we've compiled a concise summary of the podcast in this blog post. To skip ahead directly to specific sections of the blog, feel free to use the links above.
In the twelfth instalment of our AI-powered adventure, we dive deep into an array of AI and tech stories that have sparked considerable interest recently. This week, we come across an exciting mix of news items and revelations in the ever-evolving realm of artificial intelligence:
- One of the recent notable developments that could have significant implications for the European AI landscape is OpenAI's potential exit from the EU, should the proposed new AI regulations pass. OpenAI's CEO, Sam Altman, stated that if compliance isn't technically feasible, OpenAI might cease operating within the EU, which could affect both the tools reliant on OpenAI's technology and the users of ChatGPT based in the EU. However, this potential vacuum might pave the way for AI firms based in Europe to gain traction by adhering to EU regulations. We discuss this intriguing possibility and its potential impact on the AI scene in the EU.
- A16Z's recent release of its AI cannon, an abundant resource of AI knowledge, from basic introductions to advanced learning and market analysis, is another topic we delve into. It's an invaluable tool for anyone wanting to dive deeper into the AI sphere.
- Nvidia's booming stock price, spurred by higher-than-expected earnings, also comes under our spotlight. The demand for GPUs to run generative AI models might have boosted this growth, but with major companies investing in developing their own chips, we discuss what this might mean for Nvidia's future.
- ChatGPT, OpenAI's highly advanced chatbot, has now expanded its accessibility to free users. They can now browse the web or make use of the web browser plugin, an update that is both powered and branded by Bing. However, an issue that has been brought into the spotlight is the considerable energy and water consumption of these AI models. Recent research suggests that ChatGPT could have used up to 255 million liters of water in a single month, highlighting the vast quantities of energy and water necessary to run and cool the data centers. ChatGPT reportedly had around 1.7 billion visitors in April, each visitor making an average of six queries. These figures, coupled with the estimated 0.025 liters of water consumed per query, resulted in a total of about 255 million liters of water used in that month. This massive quantity vastly outpaces the average water usage of an American family, estimated at 40,000 liters per month. The reaction to these statistics has been predominantly ambivalent, with more attention being given to the technical aspects of data center operations. Nonetheless, the conversation continues, further driving the exploration of AI’s environmental impact and the exploration of alternative solutions, such as cold-climate data centers. The world of AI continues to present intriguing insights and digressions.
As always, this episode is packed with engaging discussions, fascinating insights, and a dash of humor. Despite the occasional digressions, we promise to deliver worthwhile insights into the world of AI, so stick around and enjoy the ride!
Story 1: Anthropic raises $450M to build next-gen AI assistants
AI continues to surge as an industry tour de force, especially with notable companies like Anthropic securing substantial funding to accelerate their advancements in next-generation AI systems. A primary focus of the podcast, episode 12 spotlights Anthropic's impressive Series C funding round, amounting to a robust $450 million. This financial boost significantly bolsters the company's valuation, pushing it past the $4.1 billion mark and catching the interest of influential backers such as Google, Salesforce, and Zoom.
Anthropic's primary goal is the development of safe, reliable, and honest AI systems through their innovative Claude Bay platform. Designed as a 'constitutional' AI, Claude aims to prove beneficial rather than detrimental, committed to honesty and integrity. Significantly, Zoom has also announced a partnership with Anthropic to develop customer-facing AI products, echoing Anthropic's principles of reliability, productivity, and safety. Such alliances serve to strengthen Anthropic's standing in the market, enabling it to compete effectively against other prominent players like OpenAI.
Anthropic's distinct edge is its 100,000 token context window capability, allowing for the processing of longer prompts, akin to novel-length inputs. The company has ambitious plans to build a larger next-gen model requiring colossal computing power, necessitating further funding estimated to be around £1 billion over the next 18 months. In a daring bid to expand its product range, Anthropic seeks to secure a substantial £5 billion over the next two years.
In this burgeoning market of large language models, it appears that a more diverse ecosystem is forming. For marketers, this diversification opens up a plethora of new tools powered by different AI models, each with unique strengths and applications. Anthropic's Claude Bay stands out for its constitutional framework and extensive context window, making it a distinct choice among the typically reinforcement learning through human feedback (RLHF) models.
This podcast discussion also delved into the potential of AI for content summarisation, a strength displayed by the Claude Bay model. With content summarisation, vast amounts of text can be distilled into the key facts quickly, which may hold more promise than content generation in some contexts. It underscores the potential influence of AI tools as gatekeepers of information.
However, the process of deciding what to include or exclude in a summary may still need a human touch, given the subjectivity involved in such judgment calls. Different individuals might interpret the same piece of content differently based on their personal context and experiences, a point to consider in relying on AI for summarisation tasks.
Story 2: Microsoft announces Windows Copilot, an AI 'personal assistant' for Windows 11
Microsoft is continuing its push into AI with the introduction of Windows Copilot, a new AI personal assistant for Windows 11. This new addition is being hailed as a game-changer, as it offers a centralised AI assistant that users can tap into at any point. Similar to the Bing sidebar in Edge, this assistant is available across all applications on Windows, transforming the user experience by embedding AI functionality directly into the operating system.
For the vast number of users already employing Google Apps or Microsoft Office, the introduction of an AI assistant within their operating environment is a significant move. It not only augments their existing software capabilities but also opens up new possibilities for enhanced productivity and efficiency.
Another important aspect of this announcement is the extension of chat plugins to work in Windows. This move could significantly boost the functionality of the chat GPT, making it a powerful tool to augment chat experiences within Windows with a whole host of plugins.
However, there were concerns voiced around the coherence and uniformity of the AI assistant experience across different platforms such as GitHub, Office 365, and Windows. With these various co-pilots, it's crucial that Microsoft provides a seamless user experience across its suite of tools.
The context-aware nature of the AI assistant was also praised, as it opens the door for more personalised and effective user interactions. Being aware of the specific application a user is in could lead to more tailored and proactive assistance, enhancing the overall user experience.
Another aspect of AI integration mentioned is the potential to unlock the full functionality of applications like PowerPoint, which is often underutilised by users. With the aid of the AI assistant, users may find it easier to navigate and use more advanced features of their software, increasing their productivity and efficiency.
Finally, the speakers emphasised the role of the AI assistant in creating macros in Excel, automating routine tasks, and potentially offering insights on how to improve their workflow. However, the speakers acknowledged the necessity of an informed human to make the most out of these AI capabilities.
Story 3: Google to work with Europe on stop-gap 'AI Pact'
Continuing the discussion around AI regulation, Google has announced it will cooperate with the European Union on an interim 'AI Pact'. This agreement is seen as a precursor to the full EU AI Act, which is expected to come into effect in around two years' time.
The aim of this pact is to mitigate the most significant risks associated with the rapidly evolving AI technology. This period before the AI Act is fully implemented has raised concerns among regulators, considering the speed at which changes are occurring in the AI space. Hence, this interim pact is seen as a stopgap measure.
European Commissioner Thierry Breton, and Google CEO Sundar Pichai, are in agreement about the urgency of such measures, given the potential implications and pace of AI developments. Breton has mentioned that they cannot afford to wait until the AI regulation becomes applicable and plan to work with AI partners to develop this pact ahead of the legal deadline.
While this pact is not enforceable and is largely based on good faith, it represents a key step towards establishing a framework for responsible AI development. However, there are limited details about what the pact would involve, but there has been mention of possibly labelling AI systems based on risk levels, transparency, and other important attributes.
Interestingly, there are speculations that the AI Act might be diluted slightly in terms of transparency requirements. This is due to the fact that AI explainability continues to be a significant technical challenge. This echoes the sentiments of Sam Altman, CEO of OpenAI, who had expressed concerns about potentially leaving the EU if the requirements were too stringent. However, most believe it wouldn't come to that as regulators are expected to adopt a pragmatic approach.
Marketers, and businesses in general, are encouraged to monitor these developments closely. As businesses consider leveraging AI to improve efficiency and drive creativity, it's essential to understand the regulations and frameworks that may limit certain tools or technologies. Hence, the development of this AI Pact and the forthcoming AI Act will have implications that extend beyond the realm of AI developers and into the wider business world.
Story 4: Photoshop generative AI beta launched and easy to access
Adobe has recently introduced a new beta feature to Photoshop called generative fill. This technology allows users to use natural language commands to instruct Photoshop on how to manipulate aspects of an image, such as adding or replacing elements, or expanding the size of the image and auto-filling it in a contextually aware manner.
While there is a learning curve associated with the proper use of prompts, some users have reported that impressive results can be achieved with the tool. For instance, it can intelligently select figures from an image and replace backgrounds, as well as create elements that were not present in the original image. However, some prompts may not yield the desired results and may require some fine-tuning. For instance, if a user requested a "tattoo of a lion" to be added to an image of a horse, the AI may interpret this as a person applying the tattoo, rather than the tattoo itself.
Despite these initial setbacks, the Photoshop generative fill tool has the potential to revolutionise the workflow for marketers and designers alike, empowering them to generate complex designs with little to no formal design expertise. Additionally, Adobe's new Context-Aware Expansion tool can be used to automatically and intelligently expand images, a feature that can prove invaluable in the context of creating images for blog posts and social media advertisements.
The recent release of this tool follows the publication of a research paper on a tool called Dragon from the Max Planck Institute. Although Dragon is not yet publicly available, its demo video suggests that it could introduce contextually aware image editing capabilities that are even more sophisticated than those currently offered by Photoshop. This includes the ability to realistically adjust the size of objects within an image, such as a car's wheels or a person's clothing.
Other tools, such as Canva's Magic Edit and Magic Eraser features, are also making complex image editing tasks increasingly accessible to the average user. While these tools do not offer the extensive suite of features available in Photoshop, they demonstrate the ongoing trend towards the democratisation of design.
However, despite the considerable advancements in this field, there remain limitations to what AI-based tools can achieve. For instance, while demo videos may present impressive results, actual user experience may vary significantly depending on the specifics of the image and the prompt used. The ongoing development in this area is aimed at overcoming these challenges, and as these tools become more sophisticated, the potential applications for marketers and designers continue to expand.
Story 5: Google adds generative AI to ads and image assets
In an effort to improve its advertising tools, Google has integrated generative AI into its AdWords platform. This new feature simplifies ad creation by generating relevant keywords, headlines, descriptions, images, and other assets based on the text and context of a user's preferred landing page.
The generative AI is designed to leverage Google's extensive knowledge on effective copywriting and advertising techniques to produce potentially successful combinations of ad creatives. Users can review, edit, and tailor these automated suggestions to better suit their needs, allowing AI to facilitate campaign creation.
While this update can significantly expedite workflows, it's important to note that it doesn't replace the need for a well-crafted landing page with a compelling offer. Ultimately, the quality of the landing page will heavily influence conversion rates.
There are, however, concerns about the potential shortcomings of AI-generated ad content. While AI can churn out ads quickly and efficiently, experts question whether it can truly match the skill of world-class copywriters in creating persuasive, emotion-based copy that effectively motivates and persuades potential customers.
Despite AI's potential limitations, the ability of AI tools to test multiple variations and combinations of phrases at scale is an undeniable advantage. Unlike human copywriters who can only craft a limited number of variants, AI tools can rapidly test and refine a multitude of different combinations, finding successful matches much more quickly.
Regardless of these benefits, experts suggest that a human-in-the-loop approach might still be the most effective strategy, at least initially. Even with AI-generated ad variants, the judgment of experienced professionals in selecting the most promising options can further optimise campaign outcomes.
Furthermore, while this new feature is likely to boost ad spend and competition, potentially leading to higher click costs, it also re-emphasises the importance of effective messaging and targeted advertising based on customers' interests. These are areas where human input, primary research, and exceptional copywriting skills remain invaluable.
In the end, this development could lead to an interesting cycle where the demand for superior copywriting might increase as the number of AI-generated ads rises. Ultimately, the integration of generative AI into ad creation could serve as a valuable tool in the marketing arsenal, but it's not likely to replace human expertise and creativity any time soon.
Tool of the week: Zapier and OpenAI
The innovative convergence of automation and artificial intelligence (AI) opens up a whole new world of possibilities, transforming day-to-day tasks and workflows into efficient processes. A remarkable manifestation of this is the integration of Zapier, a popular online automation tool, and OpenAI, a leading-edge AI research lab.
This combination allows for an array of automated tasks, utilizing OpenAI's generative models. With no requirement for coding skills, a multi-step Zap can be set up to monitor a Google Drive folder for new voice note files. Once a file is detected, the automation flow uploads it to OpenAI's Whisper API, a speech-to-text transcription tool. From this transcription, a ChatGPT prompt generates a blog post, which is saved in a separate Google Docs folder. A simple voice note is thus transformed into an engaging blog post with minimal human intervention.
Beyond personal applications, such seamless integration of Zapier and OpenAI proves to be a game-changer in professional settings as well. For instance, a webinar titled "How Zapier's Go-to-Market Leaders Are Using AI" unveiled fascinating use-cases demonstrating the extent of possibilities these tools unlock. One highlight included a natural language bot on Slack that fetches crucial customer information on demand, acting like a virtual sales assistant.
The benefits extend to other platforms too. By linking HubSpot, a well-known marketing, sales, and service software, with Zapier, each new lead generated can trigger a custom-made workflow. This workflow can extract relevant information to compose a personalised follow-up email, demonstrating how AI can augment sales and marketing strategies.
However, this space is still growing and certain desired integrations, such as between LinkedIn Sales Navigator and Zapier, are not available yet. But the rapid evolution of technology is promising, and we can expect to see these gaps bridged soon.
The exploration of Zapier and OpenAI together underscores the transformative potential of AI and automation in both personal and professional realms. As more platforms tap into this, we are set to witness a revolution in how we conduct our daily tasks and business operations.
Explore more with the Artificially Intelligent Marketing Podcast
This blog merely scratches the surface of the in-depth insights shared by Paul and Martin on their podcast. We heartily encourage you to listen to the complete episode available here or on your favourite podcast platform. Additionally, we provide a full directory of past episodes here, granting you access to a vast reservoir of information about AI in marketing.
The composition of this blog post was made possible with the support of AI: the initial podcast audio was transcribed utilising AI technology, while GPT-4 (ChatGPT) aided in summarising the transcript into manageable blog sections. Our BioStrata human team meticulously reviewed and edited these sections to ensure clarity and coherence. Likewise, AI technology, provided by Clipdrop, played a pivotal role in the design of the blog image.