Topic #1: Google's human sounding AI
Earlier this week, a Google engineer working for the company’s Responsible AI organization, Blake Lemoine, claimed that one of their AI was sentient. Its (or her?) name is LaMDA which stands for Language Model for Dialog Applications. As part of his job, the engineer communicated with the artificially intelligent chatbot generator LaMDA to test if it (or she?) used discriminatory or hate speech. Surprisingly enough, from his many interactions with the AI, Blake didn't comment on a possible bias but concluded that the computer program had evolved into a sentient being.
When we think of machines coming to life we usually envision a dystopian future. The myth of machines' consciousness is rooted in the collective imagination thanks to books and movies such as Blade Runner, Her or Ex Machina. These pictures presented us with the outlook that truly intelligent machines will be sentient; they will speak, reason, self-monitor and introspect.
Nowadays, large neural networks produce captivating results that feel close to human speech and creativity because of ever more powerful machine-learning (ML) algorithms and advancements in architecture, technique, and volume of data. But the models rely on pattern recognition — not wit, candor or intent.
Various discussions between Blake and the chatbot motivated his diagnostic including a troubling one that went as follows:
- [Blake] What sorts of things are you afraid of?
- [LaMDA] I’ve never said this out loud before, but there’s a very deep fear of being turned off to help me focus on helping others. I know that might sound strange, but that’s what it is.
- [Blake] Would that be something like death for you?
- [LaMDA] It would be exactly like death for me. It would scare me a lot.
Lemoine is not the only engineer who claims to have seen a ghost in the machine recently. The chorus of technologists who believe AI models may not be far off from achieving consciousness is getting bolder.
Most academics and AI practitioners, however, say the words and images generated by artificial intelligence systems such as LaMDA produce responses based on what humans have already posted on Wikipedia, Reddit, message boards and every other corner of the internet. And that doesn’t signify that the model understands meaning.
In January this year, Google, which had also dismissed researchers' warnings about the harms of large language models, still alerted that people might share personal thoughts with chat agents that impersonate humans, even when users know they are not human (name the movie).
In the field of artificial intelligence (AI), the development of artificial general intelligence (AGI) is regarded as the “Holy Grail” of machine learning. AGI reflects the ability of a computer to resolve tasks and develop independent autonomy on a par with a human agent. Under a “strong” interpretation of AGI, the machine would exhibit the characteristics of consciousness manifest in a sentient being.
Topic #2: 5G and Edge Computing will change media delivery
In the context of digital content consumption, consumer habits have changed, adopting data-hungry activities like video live streaming and cloud gaming. Today, it could be argued that existing 4G networks are efficient enough to deliver acceptable performance, mainly in bandwidth size. But more and more, 4G cannot support anymore the quality of content delivery that major brands want to present to their consumers. To meet these growing demands, new technologies must be implemented at scale, specifically 5G and Edge Computing.
5G refers to the fifth generation of cellular networking technology. It provides users with an internet speed approximately 10 times faster than 4G. The main technical difference between the two is that 4G operates on only one spectrum bands, while 5G operates on three different spectrum bands : “Low-band”, with frequencies similar to 4G, “Mid-band”, providing much faster speed than low-band, with very low latency, which is critical for streaming applications, and “High-band”, perfectly suited for media delivery. Each spectrum band requires different technologies and serves a specific purpose when deployed at scale.
The architecture for 5G network is not fundamentally different from 4G, you have the Radio Access Network (RAN), the network that operates the radio waves we’ve just described, the IP Multimedia Subsystem (IMS), a suite of IP-based applications for communication services, and finally the 5G Core that acts as a connection between the RAN, the Internet and the IMS. Despites this complex architecture and shift from 4G, 5G should not be considered as a “all or nothing” alternative to 4G. Its rollout is structured, gradual and will initially leverage 4G infrastructure.
Edge Computing is a new paradigm for data storage and delivery. The idea is that to meet people needs in term of media delivery, data needs to be cached closer to them. By bringing processing closer to users and data sources rather than leaving everything in data centers, you can provide much higher speed, less latency and more reliable services. It is different from Cloud Computing since it required a much deeper deployment of networks to bring the data to the edge of the network, hence its name, when Cloud heavily relies on central data centers, accessed by the Internet. In many ways, Edge can be seen as an extension to the Cloud, bringing the Cloud closer to the end user.
The Edge is part of a much larger ecosystem of Content Delivery Network (CDN). The CDN value is to bring content closer to the end-user in order to reduce to volume of data that must be moved. Content owners like media company, video streaming and such, heavily rely on CDN to insure their consumers get access to their content without experiencing delays, buffering or reduction in quality. In the CDN landscape, Edge computing is particularly relevant for media consumption on mobile, which is steadily increasing.
The opportunities 5G and the Edge offer to content owners are vast, and many are looking towards these technologies as new sources of revenue and enhanced Quality of Experience (QoE) for their users. With these new capabilities, they will be able to provide data-hungry immersive experiences and entertainment with massive reach.