How Much Value Can Connectivity Providers Actually Produce Using AI with Their Data?

It is inevitable that connectivity executives will talk about the ways they can employ generative AI (large language models) to support their businesses and products, ranging from the obvious (customer support or internal information queries) to some that might not be so obvious. 

While it might be easy to envision how marketing; advertising; media; content; e-commerce; education and training; financial services or healthcare could use GenAI, it likely is harder to envision how telcos can use GenAI in areas outside customer service interactions, especially in ways directly related to revenue generation. 

The standard answers might include generation of data sets that preserve the statistical properties of real customer data but are completely anonymized, then sold to customers wanting to use such data. 

At least in principle, anonymized reports could illuminate behavioral trends such as:

• Foot traffic analysis

• Traffic patterns and congestion analysis

• Device usage patterns

• App usage trends

• Network usage

• Mobile money transactions

Some forms of network performance data could be useful in some B2B situations as well. 

• Network performance insights might be useful for infrastructure suppliers

• Benchmarking of service providers and geographic areas likewise could be useful for infra providers

In practice, it is a bit unclear how valuable such data is, given that other data sources exist that also can provide insight on location; mobility; device usage; app usage or network usage. And most connectivity providers are not suppliers of mobile money services. 

Industry	Potential Value	Examples	Source
Marketing & Advertising	- Generate personalized marketing content (e.g., ad copy, social media posts) - Create targeted advertising campaigns based on audience insights - Personalize customer experiences across channels	- Generate product descriptions, headlines, and social media content tailored to specific customer segments. - Create personalized marketing materials with high engagement potential. - Develop dynamic pricing strategies based on real-time market analysis.	- Gartner
Media & Entertainment	- Generate personalized content recommendations - Create new forms of interactive entertainment experiences - Personalize news feeds and content curation	- Recommend movies, music, or articles based on individual user preferences. - Generate personalized story outlines or even short stories based on user input. - Create interactive narratives that adapt to user choices.	- Cognite
Customer Service & Support	- Automate routine customer inquiries with chatbots - Generate personalized troubleshooting guides and FAQs - Enhance customer service interactions with natural language processing	- Develop chatbots that can answer common questions and resolve basic issues. - Create personalized knowledge base articles and FAQs tailored to specific customer situations. - Improve sentiment analysis and personalize customer interactions.	- NBER
E-commerce & Retail	- Generate personalized product descriptions and recommendations - Optimize product search and discovery - Automate content creation for marketing and advertising	- Create unique and engaging product descriptions that resonate with target audiences. - Personalize product recommendations based on individual browsing behavior and purchase history. - Generate automated social media posts and marketing materials.	- Paul Prae
Education & Training	- Personalize learning experiences for individual students - Create interactive and engaging educational content - Develop adaptive learning systems that adjust to student needs	- Generate personalized study guides and practice problems tailored to individual learning styles. - Create interactive simulations and role-playing scenarios for practical learning. - Develop adaptive learning platforms that personalize content and difficulty based on student performance.	- Pro IO
Healthcare & Life Sciences	- Generate personalized reports and insights from medical data - Assist with drug discovery and development - Analyze medical literature and identify potential treatment options	- Generate personalized healthcare reports summarizing medical history and treatment options. - Analyze large datasets of medical research to identify potential drug targets. - Assist with summarizing and synthesizing medical literature for researchers.	- Investopedia
Financial Services	- Generate custom financial reports and investment strategies - Automate financial data analysis and risk assessment - Personalize financial advice and recommendations	- Create personalized financial reports summarizing investment performance and financial health. - Analyze market trends and generate investment recommendations based on individual risk tolerance. - Develop chatbots that can answer common financial questions and provide basic financial advice.	- Jon Cooke

AI Clash Between Copyright and New Technology is an Old Tale

Every new technology brings with it new legal issues. Artificial intelligence, for example, raises copyright issues. 

It is not the first time new technology has clashed with established notions of copyright. 

When photocopying machines were commercialized, manufacturers tried to block the use of the machines for making copies of copyrighted work.

Sony tried to block the use of videocassette recorders to time shift video content for later viewing. Similar disputes erupted over the use of audiocassette tapes, music file sharing and video streaming as well. 

New Technology	Copyright Issues	Key Court Decisions
Photocopying Machines (1960s)	Mass reproduction of copyrighted materials without permission.	Fair Use Doctrine Established: Williams & Wilkins Co. v. United States (1964) established the four-factor fair use test: purpose and character of use, nature of copyrighted work, amount and substantiality of portion used, and effect of use upon the market. Copying for educational purposes could be fair use.
Audio Cassette Tapes (1970s)	Home recording of copyrighted music threatened record sales.	Audio Home Recording Act (1992): Established a royalty levy on blank audiotapes to compensate copyright holders for potential lost sales due to home recording.
MP3 Players and Napster (1990s)	Peer-to-peer file sharing enabled widespread music piracy.	A&M Records v. Napster (2001): Napster was found liable for contributory copyright infringement for failing to prevent users from sharing copyrighted music.
Streaming Services (2000s-Present)	Distribution model challenged traditional music licensing and revenue streams.	Negotiated Licensing Agreements: Streaming services like Spotify and Apple Music pay licensing fees to copyright holders based on user streams.
Digital Video Recorders (DVRs)	Shifting time viewing challenged broadcasters' control over programming.	Sony Corp. v. Universal City Studios (1984): Upheld the fair use of time-shifting for personal viewing using VCRs.

Similarly, conflicts have erupted over content, social media, search, open source software and user-generated content, for example. 

Content Issue	Copyright Issues	Key Court Decisions
Social Media Sharing	Sharing copyrighted content like photos, videos, and music raises questions of fair use and infringement.	Blurred Lines: Perfect 10 v. Amazon (2002) established thumbnails could be fair use for linking purposes. However, sharing entire works without permission is generally considered infringement. The specific context and amount used determine fair use.
User-Generated Content (UGC) Platforms	Platforms like YouTube or TikTok host user-uploaded content, potentially infringing on copyrights.	DMCA "Safe Harbor": The Digital Millennium Copyright Act (DMCA) provides a safe harbor for platforms if they remove infringing content upon notification from copyright holders. Platforms like YouTube have automated takedown systems based on copyright claims.
Software Sharing and Open-Source	Sharing copyrighted software raises concerns about piracy and unauthorized distribution.	Open-Source Licenses: Open-source licenses like GPL (General Public License) allow for modification and sharing of software code, as long as certain conditions are met. These licenses provide a framework for collaborative software development while protecting copyright.
Content Aggregation Services	News aggregators like Google News display headlines and snippets of copyrighted news articles.	Fair Use and Fair Reporting: Courts have generally allowed news aggregation under fair use for purposes of reporting and commentary. The amount and substantiality of content used are crucial factors.

Eventually we will figure out some balance between copyright and use of the new technology in non-infringing ways. But it may take a while.

"Irresistible" AI Storylines Could Prove False

Some storylines seem inevitable and irresistible, either to journalists or screenwriters. And some irresistible storylines are “always wrong.”  

In the past, it has often been argued that the United States was behind, or falling behind, for use of mobile phones, smartphones, text messaging, broadband coverage, fiber to home, broadband speed or broadband price. But the “behind” storyline has proven incorrect, over time. 

To be sure, the U.S. market rarely ranks first on any “adoption or performance” metric, for a variety of reasons related to continental-sized land mass and highly rural density, among other things. 

So elections or industry standing are seen as “horse races.”  Be prepared for many surprises, as we have often seen “early leaders” falter in the computing industry. 

So artificial intelligence is viewed through the lens of a horse race or competition with a winner, a second-place finisher and many losers. So there is a “conventional wisdom” that OpenAI and Microsoft are leading, while others, such as Google, are behind, and a few, such as Apple, are “way behind.” You’d be very hard pressed to find a journalist, a financial analyst or perhaps even some within the generative AI model industry who would dispute the characterization. 

But it might also be worth noting that lots of long-gone firms were “in the lead” in the early days of the personal computer revolution. Many firms started early in the transition from character-based to multimedia web industries faltered, and many fell quickly. 

The point is that major technology transitions tend to be littered with failures, even among firms that seemed to lead, early on. Even in more prosaic industries, such as the use of mobile phones, whole countries often can be viewed as leaders or laggards in terms of consumer adoption. But such “leads” have proven to be quite temporary. 

Still, there is something different about AI. Prior transitions in computing often were led by small startups (some of which later attained scale). The current transition arguably is led by a mix of current computing leaders and some startups. That is quite different.

Is Generative AI a Form of "Graphical User Interface?"

It is unclear how widespread use of artificial intelligence, including large or small language models or generative AI might affect the human-machine interface, but in the past, a major shift in device usage has driven key changes, such as the shift from “web pages” to “mobile apps” as the primary internet access interface.

For many users, generative AI might arguably provide most value as a form of automation that enables us do things faster or better. Not often do we think of Gen AI as a form of machine interface, but that it is.

If a major shift from smartphones to other form factors (pins, watches, pendants, flexible devices) occurs, then a shift from apps to “voice interface” could happen. And that might also affect the primacy of the mobile app as well, if AI systems are able to respond without a user having to specify a particular app for use. 

User experience requirements have been crucial. Web pages designed for desktops were cumbersome on smaller mobile screens, so mobile-optimized apps offered a user-friendly interface, optimized for touch interaction with buttons, menus, and content specifically designed for the smaller screen that is assumed to be “always with the user” and available “on the go.”

Also, some app functions can be conducted without internet access, something not possible with web pages. 

Apps additionally can send real-time updates and alerts directly to users, a feature less easily supported on traditional web pages.

Designed primarily for smartphones, apps can leverage features including  GPS, camera, microphone, and sensors. And apps are tailored for specific operating systems, allowing faster performance. 

Apps also can store data locally on the device, reducing the need to constantly download information from the internet, allowing faster performance. 

App stores also allowed developers to create lighter-weight and tailored apps that might resemble “extensions” on a PC operating system.

The point is that once user engagement time shifted from the PC to the smartphone, apps became the preferred human interface. 

New device form factors without screens will offer another evolution, though perhaps not directly related to the advent of AI or generative AI capabilities. 

-------------------

Is EU "Fair Share" Dead?

European Union "fair share" proposals that would tax a few hyperscale application providers to help fund broadband upgrades by internet service providers remain uncertain. Some expect a finalized proposal by the summer of 2024, but aside from the ISPs themselves and some EU regulators, there seems little support. 

The European Commission held a public consultation on the "fair share" model in October 2023 and a majority of respondents, including academics, some regulators and most stakeholders except the large network operators who proposed the model, expressed opposition. 

In principle, the additional costs could be borne by the app firms (assuming they simply take a hit to their earnings); consumers (who will pay higher subscription fees) or business partners (advertisers, employees, other value chain partners). 

It seems doubtful the affected firms would simply take a financial hit and try and not attempt to recover the costs elsewhere. 

Some might point to South Korea, where the “sending party pays” approach was pioneered for large content suppliers. Withdrawal from the market also remains a possibility, at least for some services and apps less central to affected firm revenues. 

Twitch, the South Korean gaming service owned by Amazon, shut down at the end of 2023 and some observers say costs imposed by South Korea’s “fair share” payments by content providers to internet service providers. 

That is difficult to assess since there is no public information on the payment structure, assumed to be some form of a fee on earned revenues in the South Korea market or traffic volume or both. 

The South Korean "Fair Share" rules, under the Service Stabilization Act, only apply to a specific set of large content providers/ 

The law focuses on the five largest content providers in South Korea, based on daily user numbers and traffic share. As of 2023, these companies include Google; Netflix; Meta (Facebook, Instagram, etc.); Naver (South Korean search engine) and Kakao (South Korean messaging app). 

Those five companies are estimated to represent over 41 percent of all South Korean internet traffic.

The rules apply to firms with a minimum of one million daily users and representing at least one percent of South Korea's total internet traffic.

Small and medium content providers, including startups and individuals, are not subject to these rules. 

Before Twitch, Pandora TV went out of business because they could not afford to pay the network fees, says South Korea’s Open Net. That might not be the case, as Pandora TV suffered with revenue issues and does not appear to be covered by the law.

AI at the "Picks and Shovels" Stage

At least if one examines what sorts of startups are getting generative AI venture capital backing at the moment, though artificial intelligence remains in a “picks and shovels” stage where infrastructure creation is foremost, the next wave of applications and features are preparing to emerge, clustered in part around functions such as automation of existing business processes; data analysis; experience personalization; product development and decision support. 

source: CB Insights 

Category	Description	Examples of AI Solutions
Automation and Efficiency Improvement:	Reduce manual work, optimize processes, and increase productivity.	AI-powered robotic process automation (RPA) for repetitive tasks.  Machine learning for predictive maintenance in manufacturing.  AI-powered chatbots for customer service. Some firms already public include UIPath and Appian.
Data Analysis and Insights Generation:	Extract valuable insights from large datasets to inform decision-making.	Natural language processing (NLP) for analyzing customer sentiment and feedback.  Image recognition and computer vision for product inspection and anomaly detection.  Machine learning models for market forecasting and risk assessment. Palantir and Databricks are examples of firms in this category that already are public or at commercial scale.
Personalization and Customer Experience:	Tailor experiences and recommendations to individual user needs and preferences.	Recommendation engines powered by machine learning for e-commerce platforms.  AI-powered chatbots for personalized customer support.  Dynamic pricing models based on AI-driven customer data analysis. Criteo is a public company of this type.
Innovation and Product Development:	Accelerate new product development and enhance existing products with AI capabilities.	Generative AI for creating new product designs or marketing materials.  AI-powered drug discovery and materials science research.  Machine learning for product defect prediction and quality control. Mterialise and Autodesk are established firms of this type.
Decision-Making and Optimization:	Automate complex decisions or optimize resource allocation based on data analysis and algorithms.	AI-powered trading algorithms for financial markets.  Machine learning models for logistics and supply chain optimization.  AI-assisted medical diagnosis and treatment planning.C3AI is an example of a public company in this category.