Evaluation of my 2024 predictions

?

My predictions for 2024 covered three major themes: a worrying rise in deep fakes, the achievement of live simultaneous language translation, and the widespread use of AI companions. The predictions proved accurate in several areas, although the scope and impact of each one varied a fair bit.

?1.?????? The Year of the Deep Fake

My assertion that 2024 would see an escalation in the sophistication and prevalence of deep fakes proved to be pretty accurate. Deep fakes did indeed become a pressing issue, driven by the increasing accessibility of generative AI tools. High-profile cases of corporate blackmail and political disinformation surfaced globally, validating my concerns about the impact on elections and corporate reputations. For instance, deep fake videos reportedly influenced public opinion during the U.S. elections and created confusion in international conflicts from Ukraine to the Middle East. My prediction that instructions from a CFO could be faked came to pass in February when someone spoofed the CFO of a UK-based engineering company on a live video call, convincing a poor finance manager in Hong Kong to wire $25million to 10 bank accounts in China (https://edition.cnn.com/2024/02/04/asia/deepfake-cfo-scam-hong-kong-intl-hnk/index.html). While the adoption of AI tools to combat deep fakes has grown, for the moment the bad guys are ahead. Detection tools are available but remain unevenly deployed, often lagging behind the evolving threats. I’ll give myself a 4/5 on this one.

?

2. The Fall of Language Barriers

?My vision of significant advancements in simultaneous translation was realized, but not as quickly as I anticipated. Tools like Zoom and Teams introduced real-time transcription and rudimentary translation capabilities, while some startups achieved breakthroughs in preserving voice tone and context across languages. Multilingual virtual meetings became far more accessible, particularly in corporate settings, moving us close to my prediction of seamless communication.

However, some significant gaps remain. While tasks such as voice-to-text and text translation have improved dramatically, combining these with high-fidelity and real-time voice rendering remains challenging. Early adopters of live translation tools praised their potential but noted issues with accuracy and latency, particularly in non-mainstream languages. Moreover, adoption outside of virtual environments—such as wearable devices—has yet to become mainstream, signalling that I overestimated the timeline. While we are close to achieving live, machine-based simultaneous transformation, we’ll probably have to wait until mid-2025 to get into widespread use. I’ll take a 3/5 here.

?

3. The Rise of AI Companions

The concept of AI companions (or agents as they are now being called) materialized in 2024 but fell short of the transformative impact that I envisioned. Many generative AI platforms introduced personalized assistants that could remember user preferences and automate basic tasks, such as managing calendars or drafting emails (i.e. CoPilot). These tools, however, did not yet achieve the seamless integration or predictive capabilities that I described. While early adopters lauded the productivity gains, widespread adoption has been tempered by fears of misuse and a lack of intuitive user interfaces. Once again, I think we’ll see a major uplift in AI companions in 2025. I’ll give myself a 3.5/5.

?

Final Assessment

Overall, my predictions were perhaps a little optimistic – simultaneous translation and AI companions seem only now to be hitting their stride. While we witnessed plenty of advances in AI capabilities during the year, the maturity curve seems to be flattening as apps and models add progressively fewer new capabilities.