> It’s easy to have a culture of taking risks when you have unlimited revenue.
A company can manage to take risks with the right culture and the right business priority, even if it has limited profit. AWS, for instance, gives its services years of runway to prove themselves. The company places enormous demand on its leadership to show "technical boldness" and business initiatives. S3, EC2, Serverless, DDB, and etc, were initially considered risky bets. On the other hand, the same AWS culture does not necessarily encourage long-term applied research. As a result, AWS couldn't come up with breakthroughs like BERT and GPT. Case in point, AWS adopted a German's university's ASR solution so AWS could go to market quickly, even though the cost was that the engineers and scientists were bogged down by maintaining the model instead of coming up with models like Whisper.
A company can manage to take risks with the right culture and the right business priority, even if it has limited profit. AWS, for instance, gives its services years of runway to prove themselves. The company places enormous demand on its leadership to show "technical boldness" and business initiatives. S3, EC2, Serverless, DDB, and etc, were initially considered risky bets. On the other hand, the same AWS culture does not necessarily encourage long-term applied research. As a result, AWS couldn't come up with breakthroughs like BERT and GPT. Case in point, AWS adopted a German's university's ASR solution so AWS could go to market quickly, even though the cost was that the engineers and scientists were bogged down by maintaining the model instead of coming up with models like Whisper.