While AI offers specialized communication skills, it creates new data-storage and security challenges.
Many of us would struggle to grasp the concept of a zettabyte in any practical sense. Mathematically, it's 1 trillion gigabytes and, between we humans and our machines, we expect to generate more than 180 zettabytes of digital data in 2025. Right now, about 330 million terabytes are being introduced into the world every day – that's equivalent to the entire US population filling their OneDrive allowance on a daily basis. According to this essay on the World Economic Forum, storing our data will present major challenges: the way things stand, in 100-150 years' time there will be more data bits than atoms on the Earth and storage will consume more than the total energy generated today.
Problems notwithstanding, our prodigious output is an impressive human achievement. We have progressed through cave paintings, smoke signals, the invention of paper and books, to the many prolific techniques we have available today. It's all about the drive to communicate and express ourselves, which is embedded deeply in our nature.
Now assisted by AI, we could exceed even our own predictions. With more or less universal access to large-language models like GPT (generative pre-trained transformer), we can produce text at an incredibly fast rate. Text-to-image models like OpenAI's DALL-E and Google's Imagen now provide even more powerful tools for creative expression. We can create digital artwork that most of us simply would not have the skills to produce any other way, and in vastly less time than would be needed using traditional media or even manual digital tools. This is incredibly empowering technology that can be used to share ideas, visions and creative passions in more natural and intuitive ways than previously possible. On the other hand, the data mountain is set to grow exponentially.
There will also be transformative effects on our ability to communicate from person to person in real time, in the real world. Soon, we could be able to translate directly between languages, when talking face-to-face, with no discernible latency. By promoting interactions across language and cultural barriers, this could greatly enhance mutual understanding and realize numerous benefits.
On the other hand, while the extent, volume and importance of our digital communications increases with the assistance of AI, the same technology is also putting more power in the hands of fraudsters. While AI can help us to express ourselves, it can help others to overcome the authentication mechanisms that protect our personal content and our identities on social platforms, financial services, healthcare, and other Internet-based services.
Some recent scams have sent texts to victims, claiming to be a friend in trouble and seeking assistance, usually involving a money transfer to bail them out of a desperate situation. Some wary targets have had the presence of mind to call their friends for confirmation. But a deepfake capable of replicating voice or appearance would be much more difficult to detect. Then imagine that deepfake further augmented with recognizable personal characteristics such as mannerisms or idiosyncrasies like favorite phrases or even words that we may misuse or misspell. Recall that spooky scene in the first Terminator movie, when the cyborg replicates a female voice to get information during a phone call; now modernized and more powerful. We may need the help of defensive AI to spot the anomalies.
A US study of identity fraud by Javelin Strategy and Research found there were 15 million US victims in 2023, losing $23 billion to scams alone. Since the pandemic, online identity fraud has increased significantly as scammers joined the work-from-home revolution, creating fake online accounts using stolen identities and taking over existing accounts. Strengthening authentication mechanisms using techniques such as two-factor identification and biometrics is the necessary next step in what will certainly become an escalating battle.
Deepfake has the potential to overcome biometric authentication, including fingerprints, retinal scans and behavioral biometrics. According to security experts, combining multiple authentication mechanisms can strengthen protection. Token- or certificate-based authentication can be more resilient against stolen credentials. On the other hand, these are more complicated to implement and gaining access can take longer. The ease of use that goes hand in hand with the convenience and immediacy we enjoy needs to be preserved.
While the battle against AI-powered fraud will continue, data centers face their own battle to handle the increasing intensity of AI workloads. Photonics may be able to help, and startups are already finding ways to create the faster and more efficient modulators that will be needed. Optical networking is already reaching ordinary homes, but here the opportunity is to use photonics at the server level to move data at extremely high speed between processors, and even within the chips themselves, to utilize the computing resources available throughout the data center to maximum effect.
Board-level photonics at these data speeds is an esoteric technology that's consuming significant startup capital. It could quickly become routinely adopted in our quest to manage the huge amounts of bits that make up the data we are now seeking to collect, process, create and store in every minute of every day. Ultimately, our goal is to communicate – the process of sharing information, knowledge, emotions, experiences, feelings, helping us all increase our understanding of ourselves, each other and the world around us.
Alun Morgan is technology ambassador at Ventec International Group (venteclaminates.com); This email address is being protected from spambots. You need JavaScript enabled to view it.. His column runs monthly.
