Software giant embraces AI: How Adobe completes the "AI transformation" step by step

With the rise of AI, various industries are undergoing tremendous changes. Software giant Adobe has seized the opportunity presented by artificial intelligence and depicted a model for the transformation of the AI era.

Three months ago, when the AIGC platform Midjourney created lifelike portraits, the market believed it was the end of Adobe.

However, in the face of the challenge posed by AI, this leading creative software company made a decisive and swift turn. AI not only did not hinder Adobe's business but also became a catalyst for its revenue growth.

How did Adobe transform itself from an "AI loser" to an "AI winner" and successfully complete its "AI transformation"?

Seizing the main vulnerability of AI - copyright security

1. A meeting that determined market positioning

Nine months ago, Adobe's Chief Strategy Officer, Scott Belsky, sat in a meeting room in New York with over twenty top marketing executives to discuss how artificial intelligence could change advertising and enable them to turn a series of words into a visual advertisement.

Initially, Belsky believed that they were in agreement regarding the immense potential of AI. However, the marketers insisted that they were unwilling to use AI image models trained on copyrighted content, as this could lead to collective lawsuits.

This conversation helped Belsky and Adobe establish their position in the competitive landscape of generative AI.

This software giant, valued at $219 billion, chose to gradually and steadily integrate GAI tools into its core design products such as Photoshop, Illustrator, and Premiere. They believed that a cautious approach was necessary and did not want to take the risk of using AI models that could potentially trigger collective lawsuits.

In recent months, Adobe has been steadily accelerating its pace. In March, Adobe released a beta version of its AI tool, Firefly, which trained on hundreds of millions of images from the company's vast image library, as well as some publicly licensed and copyright-expired images from the public domain. Last month, Adobe integrated Firefly's functionality into the beta version of Photoshop.

Since the model was trained on licensed and public domain photos, marketers told Belsky that they would now consider using it for commercial content. Two weeks ago, Adobe announced that it would provide intellectual property compensation to customers, safeguarding them against copyright infringement lawsuits resulting from images generated by Firefly.

**Firefly is currently in the testing phase, but Belsky said that the company will be able to use it commercially "soon". Adobe is currently working with Japanese marketing agencies, IBM, and Medtronic to explore how they can use Firefly in their work.

Adobe aims to be the white knight in the AI competition, contrasting with Stability AI's diffusion model, OpenAI's Dall-E 2, and Midjourney, which scrape millions of images from the internet to train their models, disregarding copyright laws aimed at protecting artists and their works.

2. Key Step: Acquiring the Design Newcomer Figma

Scott Belsky, Chief Strategy Officer of Adobe, was previously the founder and CEO of the Behance design community, which aimed to protect the rights of creatives. Adobe previously acquired Behance for $150 million, and Belsky took on the role of Chief Strategy Officer and EVP of Design and Emerging Products at Adobe in March of this year.

Over the past few years, Belsky has made his mark throughout Silicon Valley, serving as an angel investor for companies such as Pinterest, Eight Sleep, Warby Parker, Uber, and Airtable. Leveraging his entrepreneurial connections, Belsky helped secure Adobe's largest acquisition to date - a $20 billion acquisition of the graphic design company Figma. Belsky and Figma's co-founder Dylan Field have been friends for many years.

However, this deal is now facing scrutiny from the US Department of Justice, which is reportedly preparing to file an antitrust lawsuit to block the transaction. With its $20 billion Figma acquisition at risk, the software giant is shifting its focus to Firefly, a generative AI model with copyright protection capabilities.

Belsky is working together with Adobe's Creative Cloud, Research, and Digital Media teams, hoping that Adobe can protect creatives, many of whom are seeking compensation for using their works to train AI models. Currently, there are several copyright infringement lawsuits against Stability AI and Midjourney, and there is still no legal consensus on what constitutes fair use of AI in the courts.

3. Robust "Transformation" to Protect Copyright

In fact, Adobe's work on generative AI began in 2019 when Alexandru Costin, Vice President of AI at Adobe, demonstrated how AI can tilt the Mona Lisa's mouth into a smile or change her pose. When AI technology was still far behind Costin's vision. He said that no one could see how this blurry 64-pixel image would become a fully editable 4000-pixel image two years later. Belsky advised him to be patient, as the technology was not yet good enough to be integrated into Adobe's flagship products, but its time would come soon. However, Adobe was already behind its competitors when it launched the Firefly tool in March.

Nevertheless, Belsky said that users' interest in Firefly even exceeded Adobe's highest expectations, as they created 500 million images using the tool in the three months after its launch.

Importantly, although companies like OpenAI, Stability AI, Midjourney, and Runway may be pioneers, Adobe believes it has a crucial advantage: after watching other companies blatantly infringe copyrights and face lawsuits, the company can now seize the market dominance by offering more trustworthy products.

4. Will countries stand with tech companies?

AI infringement has become an urgent issue, according to David Hoppe, a lawyer at the San Francisco-based Gamma Law Firm:

This situation is rapidly evolving, and a series of copyright infringement lawsuits against Stability AI and Midjourney - from Getty Images and collective lawsuits from artists - could have a huge impact in the field of AI.

Intellectual property lawyers have two concerns about generative AI models: first, whether any images created by AI models can be protected by copyright; second, whether these models violate copyright laws by stitching together images from the internet.

So far, artists have not won this debate, and if Stability AI and Midjourney win their lawsuits, their models will continue to generate new images using existing artworks. In that case, it will be difficult for Adobe's Firefly model to dominate.

Belsky bets that other AI tools will eventually have to consider strict copyright restrictions, and many will follow the path of music streaming service Napster.

The initial music streaming emerged as a cultural force at the turn of the millennium, when it was a moment of free competition, without legislation or real rules and structure. It collapsed under the impact of lawsuits from the record industry, opening the door for new companies like Spotify, which came with licensing agreements from record companies.

Firefly is the way Adobe jumps to the end of the Napster story, and it has released a legally defensible and "ethical" version since the beginning of the rise of generative AI.

Belsky said that in the future, Adobe can reach licensing agreements with entertainment companies, allowing Firefly users to use well-known characters or brands in their works, just like the collaboration between YouTube and Spotify with music companies.

Currently, Adobe's efforts to respect copyright are risky. Last year, the UK government announced that it would allow AI developers to train on copyrighted content. In response to the music industry's backlash, it retracted this statement in March and reconsidered AI regulations in the UK. Soon, US laws may also stand on the side of large model companies.

Morgan Stanley: IT Service Providers to Benefit from the AI Wave

For IT service providers embracing AI, Morgan Stanley believes that in the short term, they may be beneficiaries of data and AI-driven spending.

Morgan Stanley points out that advances in deep learning applications, such as generative AI, unlock various new jobs that can be automated. While previous technological advancements increased or automated routine, rule-based tasks, AI will impact more complex tasks that require cognition, learning, and decision-making. The automatic generation of natural language, code, and numerical content, as well as task execution, could drive significant efficiency gains in professional services, including the over $1 trillion IT services industry.

Morgan Stanley sees IT service providers as beneficiaries of AI in the near term as a whole, believing that early adopters of AI tools internally can benefit from productivity and marginal benefits. Morgan Stanley expects this scope to be limited to early adopters who can enhance user productivity and strategically invest in building AI capabilities for IT service providers.