58:00 - Block chains can't out perform Tx/sec of Visa w/o losing the trustless attribute. I will challenge him on that with my white paper and argue that we can't have the trustless attribute w/o scaling!
They don't need to outperform them. Blockchains only need to get the job done.
Human transaction needs are finite. Meaning, that even if technology, whether centralized or decentralized, can process 10 trillion tx/sec, it's useless to us because we have, say, 10.000 tx/sec in terms of actual needs.
Now, technology went 1000x from 1995 to 2015, in terms of cpu, storage, ram, network bandwidth.
A 486 or Pentium PC, with 4-16mb ram, 1gb disk and 28kbps modem is now replaced with 1000x+ more powerful processor, gbytes of ram, terabytes of disk and mbps of network connection.
If we go another 1000x to 2035 and if we go yet another 1000x to 2055, then the current 5tx/s would be 5.000 and 5.000.000 in 2035 and 2055 respectively. And that's not accounting for
a) software optimizations
b) software exploiting the hardware better (SIMDs, GPUs etc)
So what we think of blockchains changes depending the time coordinates. If blockchains existed in their current form back in 1995 they would be a nice but useless thought experiment. In 2015 they are just starting to work in terms of hardware and in 2035 they'll probably be pulling more throughput than (current) visa - but even if visa can do 1000x by then it won't matter because there won't be 1000x more txs for them to handle due to finite consumer needs.
If the capabilities of the decentralized network scale more than performance increases in terms of hardware or software, then centralization occurs as a result to more specialized equipment and data centers.
If the capabilities of the decentralized network scale in sync with increases of hardware and software speed, then there is not much loss of decentralization. To give an example, if a 2035 pc and home connection can sustain a 5k tx/sec network with its storage, processing and bandwidth requirements, just like a 2015 pc can sustain a 5tx/sec network with its respective requirements, then there is no issue in terms of decentralization.
The above doesn't mean we must wait until 2035, or wait for hardware. There is work to be done in the software optimization level.
edit: fixed / thanks Smooth.