On Tuesday, January 25th, Philip Rosedale held a Twitter Spaces event (also relayed in-world in Second Life and to other virtual spaces) to discuss “the metaverse” and Second Life and to answer questions from SL users / interested parties. He was joined from time-to-time by various guests notably Avi Bar-Zeev¹ who added their own thoughts.
The conversation was wide-ranging, extended over some 100 minutes. What follows here is an attempt at a summary of the key areas of discussion in terms of comments from Both Philip Rosedale and Avi Bar-Zeev. Given the natural flow of the event, with subjects being raised and returned to, rather than being discussed sequentially, I have attempted to summarise the comments into bullet-point under topic headings. Where I have felt it worthwhile, I have included audio extracts of the actual comments made as well.
General Notes for This Summary
- The full audio for the event is currently available via Twitter Spaces, where it will remain available for 30 days from the date of recording – Tuesday, January 25th, 2022.
- For the most part, the bullet points refer to comments made by Philip Rosedale. Those by Avi Bar-Zeev are intentionally under a sub-heading for easier identification.
- Where provided, audio extracts below have been edited to remove pauses, repetition, non-relevant asides, etc., in an attempt to assist with understanding the flow of comments. Where this has been done, I have taken care to try to ensure none of the original context / meaning of the comments has been lost or in any way altered.
On His Role at Linden Lab
- Reiterated that he is not back at the Lab in any form of a managerial role or full-time “at this time” [which I found a potentially interesting qualifier, if intended that way].
- He is “delightedly” providing advice and attending meetings with Lab staff.
On Second Life’s History
- Recalled SL’s modest beginnings as LindenWorld, and interactions with the first residents – such as Stellar Sunshine.
- Noted that a challenge Second life has had throughout its history is that in allowing user-generated content (UGC) that interacts with the native controls / capabilities (such as the physics engine), it becomes increasingly hard to make substantive changes to the behaviours of those capabilities, lest they result in content becoming “broken”.
- Also noted that SL was very much ahead of its time with things like its particle and water system (the latter of which allowed for splashes, etc., when object cross the water plane) had to be removed, because they were simply too computationally complex for either home computers to process in a timely manner or the available network bandwidth and server communications could transmit to other users in a timely fashion – with some of these problems still existing to this day.
- Indicated that these are issues not confined just to SL: they are lesson that need to be understood in building any user-driven virtual spaces.
On Moderating Virtual Spaces
- Sees moderation of virtual spaces / virtual worlds as something that still needs to be fully addressed.
- Believes the approaches to moderation taken by social media platforms and across the Internet as a whole are insufficient for immersive spaces utilising avatars – simply put, a single standard of rules applied from above by a single company will not work.
- In particular sees a top-down approach to moderation troublesome for a number of reasons, including:
- Those utilising Meta’s suggested approach of recording interactions so that in the event of a dispute / reported abuse, the last 10-15 minutes can be attached to an abuse report, could use the gathered data to also help drive any advert / content-based revenue generation model they might also use.
- Top-down approaches risk utilising a “one size fits all” approach to disputes in order to minimise the costs involved in managing moderation activities, thus removing the opportunity for for subtlety of approach or taking into consideration the uniqueness of any given situation / group, potentially alienating groups or activities.
- Instead, believes that there should be a more fluid approach to moderation more in keeping with the physical world, and adjusted by circumstance / situation, and that:
- Companies need to look at how spaces within their platforms are used and what is deemed as acceptable behaviour by the people operating / using them.
- Enable the communities / groups using spaces to be able to self-moderate through the provision of the means for them to do so (e.g. provide their own guidelines backed by the ability for them to remove troublemakers).
- Recognise the fact that the majority of people will adjust their behaviour to suit the environment they are within and self-moderate according to expectations of that environment.
- Toward the end of the session, notes that there is a risk associated with some aspects of decentralisation of moderation / control. Within Second Life, for example, decentralisation of land ownership brought with it issues of anti-social behaviour and griefing – ad farms, intentionally being abusive towards neighbours through the use of large billboards, sounds, etc., whilst making the land too expensive for it to be reasonably purchased.
From Avi Bar-Zeev
Also notes that there is an inherent danger in how a company could use the recording / surveillance approach to moderation to profile users and to assist their ad / targeting revenue model.
- However, he thinks the larger issue is that given the review of recordings associated with abuse reports that may be coming in by the thousand in a large-scale system is going to be human-intensive, then the use of AI systems to manage this process and minimise costs is likely inevitable. But:
- How do we know the AI isn’t by its very nature, pre-disposed to “find bad behaviour”, and to do so without consideration of a wider context (pre Philip Rosedale’s warning).
- How can we be sure AI programming is sufficient for a system to correctly recognise some behaviour types as being abusive.
- Is dealing with incidents in retrospect and with limited supporting data (e.g. just 10 minutes of audio) actually the best method of handling incidents.
- As such, also believes it is better to design systems wherein people are innately aware that they are dealing with other people across the screen from them, and so they self-moderate their behaviour (as most of us naturally do most of the time when engaging with others), and that there are ramifications if we then chose to be directly abusive towards others. In short, virtual spaces should “re-humanise” our interactions with others.
On Preferred Business Models for Virtual Spaces
- The common practice for social platforms – YouTube, Facebook, each is the behavioural surveillance model noted above – collecting data on user interests and activities, etc., and using that to push content / adverts / etc., to users whilst also gathering an overall profile on them.
- Sees the development of AI / intuitive algorithms in this space particularly dangerous as they grow increasingly capable of recording moods / states of mind / health conditions (particularly where facial / body tracking is utilised).
- Much prefers the model offered by the likes of Second Life, where the emphasis is not on advertising revenues, content delivery for brands, etc., but rather entirely fee-based.
- Notes that as it is, Second Life generates more revenue dollars per user per year than You Tube through its model, and probably than Facebook. As such, and with roughly one million active users, SL has proven the fee-based revenue model works, and it is fully scalable.
From Avi Bar-Zeev
- Notes there has been criticism of some platforms that deal in virtual “land” than is really just vapourware.
- Wanted to underscore the point that SL and platform like it do not fall into this category, because while the land is virtual, it is nevertheless underpinned by actual servers and infrastructure and support services that incur costs that are being met by the fees charged.
On Accessibility for Virtual Spaces
- Points out that when people in Second Life talk about “accessibility”, it is invariably from the perspective of learning to do things within the platform – getting to grips with the viewer, walking, talking, building, etc., and the “steep learning curve”. However, would argue that the issue starts much earlier than that.
- The real issue with accessibility is not what to do / how to do it, but in getting people comfortable with the idea of using avatars and virtual spaces.
- Has personal experience of this through building both Second Life and High Fidelity² and notes that by-and-large a typical reaction of anyone being asked to sit down and try any virtual world / space for social interaction will likely express interest in the experience, but discomfort at the idea of making it a part of there daily interactions in the manner promoted by the likes of Meta, etc.
- Ergo, the first step in accessibility is moving things to a point where people are comfortable within idea of using avatars and a virtual presence. Only when this has been addressed, and people are comfortable with the idea, can the wider issues of moderation, world-building, economics, etc., be tackled.
- Believes the way to do this is to make avatars more visually expressive – which is itself a tough proposition [see, for one thing, the issue of the Uncanny Valley], and towards the end of the video expresses how this could be done by using webcams on laptops, mobile devices to capture facial expressions and have the back-end software then translate these onto avatar faces [an approach LL have indicated they plan to develop in 2022].
- Does see spatial audio of the kind High Fidelity has been developing as a factor in enabling greater depth of interaction, particularly within groups of people, but really sees the ability to mimic facial expressions, gestures, etc., to provide that underpinning level of non-verbal communications as a core part of making avatar-based interactions more acceptable to a larger audience.
- In terms of avatars, expressiveness, etc., does point out that avatars should not be equated necessarily to “digital twins” – that your avatar must be a digital representation of Second Life, and his opinion is that this should preferably be true in future virtual worlds / spaces.
- However [and assuming adoption of virtual spaces into the work medium] sees a possible issues over “class distinction” between those ability interact “in real life” in person or through mediums like Zoom, etc., and those interacting through the purely digital, which may have to be addressed.
On the Linden Dollar and Crypto-Currencies
- Offers a background on the Linden Dollar and why it uniquely works as a virtual currency, presenting something of a mix between crypto and regulated fiat money.
- Highlights some of the issues with current crypto and why it is presently not a good medium for virtual economies.
On Mobile – Second Life and in General
- Second Life (and Facebook) arrived before the first of large-screen, images / graphics capabilities arrived on the market in the form of the iPhone. As such, SL was solely geared towards desktop systems, as there was no reason to even consider the idea of compact, powerful mobile devices.
- Admires the way the Minecraft has made in-world building so intuitive on mobile, and believed that is something virtual worlds need to achieve.
- Personally believes it is essential for virtual worlds to offer convenient access from multiple devices, noting that perhaps the biggest world-wide platform in this regard is probably Android.
- Thus the question is one of what features can be included with a mobile solution, and which features should be included when compared to the more immersive “hands-on” capabilities.
- Allowing for his status as an advisor, he can say that Linden Lab is actively working on mobile. [I try to provide updates on this when there is news, using the SL Mobile tag in this blog.]
- Suggests that LL could possibly engage in some form of “smaller” acquisition³ or building on an open-source tool.
- In discussing the 20th anniversary of the rezzing of the first prim in LindenWorld – see: Happy 20th rezday to Second Life’s humble Prim!, noted that a good part of the magic of early virtual worlds was that of in-world, real-time building, including doing so collaboratively, and helped build a sense of social engagement and sharing which more recently platforms (or SL through mesh) have either never had or have perhaps lost.
- In talking about the primitive system, drew a comparison with the current hype around NFTs, noting that (with the introduction of the permissions system) every prim in SL is unique in terms of its creator, data and time of creation, UUID and what subsequent owner might do with it (modify it, copy it, pass it on / sell it to someone else), all of which are indelibly recorded in its metadata.
- Noted that if “the metaverse” is to be as influential on live and work, etc., as the world Wide Web, then it not only needs people, but content. In particular notes that at its peak growth, the WWW was adding 300,000 new pages of content a day (2012). Clearly, in terms of virtual spaces, an exponential growth rate is liable to prove too much for a single corporate entity to manage.
- Re-iterates the view that in terms of VR headsets, it is not the weight, the nausea or (in the case of Second Life) potential issues around frame rates, etc., that is key to increasing general adoption by consumers. Rather, it is in making the use of such headsets more inherently “safe” and less anti-social in terms of using them in physical rooms / locations where others are present.
- Avi Bar-Zeev has been a pioneer, architect and advisor in Spatial Computing (AR/VR/MR/XR) for nearly 30 years, behind the scenes in the world’s largest tech companies and at large. In early 2010, he helped found and invent the HoloLens at Microsoft, developing the first prototypes, demos, patents, plans and UX concepts, sufficient to convince his leadership. At Bing, he built first prototypes for developer-facing aspects of AR, sometimes called the “AR cloud.” At Amazon, he helped create PrimeAir as well as Echo Frames. He most recently helped Apple advance its own undisclosed projects. In 1999-2001, he co-founded Keyhole, the company behind Google Earth, and helped define Second Life’s core technology (and created the code that gives us prims). Back in the 1990s, he worked on novel VR experiences for Disney, including “Aladdin’s Magic Carpet” VR Ride, the “Virtual Jungle Cruise” and “Cyberspace Mountain.”
- For those who may not be familiar with it, High Fidelity Inc was originally set-up to create a VR headset-centric, decentralised virtual spaces / virtual world platform. However, the company pivoted away from this in 2019/2020 with the realisation that consumer VR systems are these not yet a comfortable proposition for the majority of people.
- This should probably not be conflated with any idea of buying Lumiya (which has been a constantly-stated view by some users). so far as I’m aware, there is no line of contact between Linden Lab and Lumiya’s developer.